Academia.eduAcademia.edu

EViews Illustrated for Version 8

Profile image of Atilla GökçeAtilla Gökçe

Richard Startz, University of California, Santa Barbara

descriptionSee full PDF

Abstract
sparkles

AI

EViews Illustrated for Version 8 serves as a comprehensive guide for users looking to understand the functionalities of EViews software. It begins with a quick overview of essential features, including data handling, statistical modeling, and forecasting. The text emphasizes user engagement through practical examples and step-by-step instructions, aiding both novice and seasoned users in maximizing the software's capabilities.

Figures (400)
Workfile: The Basic EViews Document You and I are going to start our conversation by taking a quick walk through some of EViews’ most used features. To have a concrete example to work through, we’re going to take a look at the volume of trade on the New York Stock Exchange. We’ll view the data as a set of numbers on a spreadsheet and as a graph over time. We’ll look at summary statis- tics such as mean and median together with a histogram. Then we’ll build a simple regres- sion model and use it for forecasting. Being impatient to get started, let’s take the quick solution and load an existing workfile. If you're working on the computer while reading, you may want to load the workfile “nysevol. ume.wfl” by clicking on the Open an existing EViews workfile button. If you have saved the workfile on your computer, navigate to its location and open it.
Workfile: The Basic EViews Document You and I are going to start our conversation by taking a quick walk through some of EViews’ most used features. To have a concrete example to work through, we’re going to take a look at the volume of trade on the New York Stock Exchange. We’ll view the data as a set of numbers on a spreadsheet and as a graph over time. We’ll look at summary statis- tics such as mean and median together with a histogram. Then we’ll build a simple regres- sion model and use it for forecasting. Being impatient to get started, let’s take the quick solution and load an existing workfile. If you're working on the computer while reading, you may want to load the workfile “nysevol. ume.wfl” by clicking on the Open an existing EViews workfile button. If you have saved the workfile on your computer, navigate to its location and open it.
Another way to think about data is by looking at summary statistics. In fact, some kind of summary sta- tistic is pretty much necessary in this kind of situation; we can’t hope to learn much from staring at raw data when we have 400 plus numbers. To look at summary sta- tistics for a series press the button and choose Descriptive Statistics & Tests/Histogram and Stats. Viewing an individual series—5
Another way to think about data is by looking at summary statistics. In fact, some kind of summary sta- tistic is pretty much necessary in this kind of situation; we can’t hope to learn much from staring at raw data when we have 400 plus numbers. To look at summary sta- tistics for a series press the button and choose Descriptive Statistics & Tests/Histogram and Stats. Viewing an individual series—5
We ve learned that in the early 21st cen- tury NYSE volume is many orders of magnitude greater than it was at the close of the 19th. In retrospect, the picture isn’t surprising given how much the economy has grown over this period. Trading volume has grown enormously over more than a century; it’s grown so much, in fact, that numbers from the early years are barely visible on the graph. Let’s try a couple of different approaches to getting a clearer picture. over time. Underlying this idea of growth over time is some sense that the financial sector o the economy is far larger than it was in the past.
We ve learned that in the early 21st cen- tury NYSE volume is many orders of magnitude greater than it was at the close of the 19th. In retrospect, the picture isn’t surprising given how much the economy has grown over this period. Trading volume has grown enormously over more than a century; it’s grown so much, in fact, that numbers from the early years are barely visible on the graph. Let’s try a couple of different approaches to getting a clearer picture. over time. Underlying this idea of growth over time is some sense that the financial sector o the economy is far larger than it was in the past.
As a first pass, we’ll look at only the last three years or so of data. To limit what we see to this period, we need to change the sample.
As a first pass, we’ll look at only the last three years or so of data. To limit what we see to this period, we need to change the sample.
Hint: Ranges of dates in EViews are specified in pairs, so “2001q1 @last” means all dates starting with the first quarter of 2001 and ending with the last date in the work- file. EViews accepts a variety of conventions for writing a particular date. “2001:1” means the first period of 2001 and “2001q1” more specifically means the first quarter of 2001. Since the periods in our data are quarterly, the two are equivalent.
Hint: Ranges of dates in EViews are specified in pairs, so “2001q1 @last” means all dates starting with the first quarter of 2001 and ending with the last date in the work- file. EViews accepts a variety of conventions for writing a particular date. “2001:1” means the first period of 2001 and “2001q1” more specifically means the first quarter of 2001. Since the periods in our data are quarterly, the two are equivalent.
the dates, and the special code “@last” to pick up the last date available in the workfile When you’ve changed the Sample dialog as shown in the second figure, hit |_°%* |.
the dates, and the special code “@last” to pick up the last date available in the workfile When you’ve changed the Sample dialog as shown in the second figure, hit |_°%* |.
Let’s turn to another approach to thinking about trading volume. Our first line graph (before we shortened the sample) presented a picture which looks a lot like exponential growth over time. A standard trick for dealing with exponential srowth is to look at the logarithm of a variable, relying on the identity y = we logy = gt. In order to look at the trend in the log instead of in the level, we’ll create a new variable named LOGVOL which equals the log of VOLUME. This can be done either with a dialog or by typing a command. We’ll do the former first. Choose the menu item Quick/Generate Series... to bring up Generate Series by Equ Enter equation logvol=log(volume) Sample 200191 200494 (ex a dialog box. In the upper field, type “logvol = log(volume)”. Notice the sample is still set to use only the 21st century part of our data. T in a moment.
Let’s turn to another approach to thinking about trading volume. Our first line graph (before we shortened the sample) presented a picture which looks a lot like exponential growth over time. A standard trick for dealing with exponential srowth is to look at the logarithm of a variable, relying on the identity y = we logy = gt. In order to look at the trend in the log instead of in the level, we’ll create a new variable named LOGVOL which equals the log of VOLUME. This can be done either with a dialog or by typing a command. We’ll do the former first. Choose the menu item Quick/Generate Series... to bring up Generate Series by Equ Enter equation logvol=log(volume) Sample 200191 200494 (ex a dialog box. In the upper field, type “logvol = log(volume)”. Notice the sample is still set to use only the 21st century part of our data. T in a moment.
Since we're trying to look at all the available data, we want to change the sample to include, well the whole sample. One way to do this is to use the menu selection Quick/Sample.... Another way to change the sample is to double-click on the sample line in the upper pane of the workfile window—right where the arrow’s pointing in the picture on the right.
Since we're trying to look at all the available data, we want to change the sample to include, well the whole sample. One way to do this is to use the menu selection Quick/Sample.... Another way to change the sample is to double-click on the sample line in the upper pane of the workfile window—right where the arrow’s pointing in the picture on the right.
Historical hint: Ever wonder why so many computer commands are limited to four let: ters? Back in the early days of computing, several widely used computers stored char- acters “four bytes to the word.” It was convenient to manipulate data a “word” at a time. Hence the four letter limit and commands spelled like “smpl”.
Historical hint: Ever wonder why so many computer commands are limited to four let: ters? Back in the early days of computing, several widely used computers stored char- acters “four bytes to the word.” It was convenient to manipulate data a “word” at a time. Hence the four letter limit and commands spelled like “smpl”.
Hint: Menu items, both in the menu bar at the top of the screen and menus chosen from the button bar, change to reflect the contents of the currently active window. If the menu items differ from those you expect to see, the odds are that you aren’t look- ing at the active window. Click on the desired window to be sure it’s the active win- dow. Looking at a pair of series together—11
Hint: Menu items, both in the menu bar at the top of the screen and menus chosen from the button bar, change to reflect the contents of the currently active window. If the menu items differ from those you expect to see, the odds are that you aren’t look- ing at the active window. Click on the desired window to be sure it’s the active win- dow. Looking at a pair of series together—11
To get started we need to create the variable ¢. In the command pane at the top of tk EViews screen type: The group shows time and log volume, that is, the series T and LOGVOL, together. Just as there are multiple ways to view a series, there are also a number of group views. Here’s the spreadsheet view.
To get started we need to create the variable ¢. In the command pane at the top of tk EViews screen type: The group shows time and log volume, that is, the series T and LOGVOL, together. Just as there are multiple ways to view a series, there are also a number of group views. Here’s the spreadsheet view.
Since we want to think about how volume behaves over time, we want to look at the variables T and LOGVOL together. In EViews a collection of series dealt with together is called a Group. To create a group including T and LOGVOL, first click on (At. Now, while holding down the Ctrl-Key, click on 4 logvol. Then right-click highlighting Open, bringing up the context menu as shown and choose as Group.
Since we want to think about how volume behaves over time, we want to look at the variables T and LOGVOL together. In EViews a collection of series dealt with together is called a Group. To create a group including T and LOGVOL, first click on (At. Now, while holding down the Ctrl-Key, click on 4 logvol. Then right-click highlighting Open, bringing up the context menu as shown and choose as Group.
The equation for the plotted line can be written algebraically as y = @+ B t. @ is the inter- cept estimated by the computer and B is the estimated slope. Just looking at the plot, we can see that the intercept is roughly -2.5. When t = 400, LOGVOL looks to be about 4. Reaching back—possibly to junior high school—for the formula for the slope gives us an approximation for B . Estimating your first regression in EViews—13
The equation for the plotted line can be written algebraically as y = @+ B t. @ is the inter- cept estimated by the computer and B is the estimated slope. Just looking at the plot, we can see that the intercept is roughly -2.5. When t = 400, LOGVOL looks to be about 4. Reaching back—possibly to junior high school—for the formula for the slope gives us an approximation for B . Estimating your first regression in EViews—13
Hint: Sometimes the dependent variable is called the “left-hand side” variable and the independent variables are called the “right-hand side” variables. The terminology reflects the convention that the dependent variable is written to the left of the equal sign and the independent variables appear to the right, as, for example, in log(volume,) = a+ Bt.
Hint: Sometimes the dependent variable is called the “left-hand side” variable and the independent variables are called the “right-hand side” variables. The terminology reflects the convention that the dependent variable is written to the left of the equal sign and the independent variables appear to the right, as, for example, in log(volume,) = a+ Bt.
Having seen the picture of the scatter diagram on page 13, we know this line does a decent job of summarizing log( volume) over more than a century. On the other hand, it’s not true that in each and every quarter LOGVOL equals — 2.629649 + 0.017278t, which is what the equation suggests. In some quarters volume was higher and in some the volume was lower. In regression analysis the amount by which the right-hand side of the equation misses the dependent variable is called the residual. Calling the residual e (“e” stands for “error”) we can write an equation that really is valid in each and every quarter: Estimating your first regression in EViews—15
Having seen the picture of the scatter diagram on page 13, we know this line does a decent job of summarizing log( volume) over more than a century. On the other hand, it’s not true that in each and every quarter LOGVOL equals — 2.629649 + 0.017278t, which is what the equation suggests. In some quarters volume was higher and in some the volume was lower. In regression analysis the amount by which the right-hand side of the equation misses the dependent variable is called the residual. Calling the residual e (“e” stands for “error”) we can write an equation that really is valid in each and every quarter: Estimating your first regression in EViews—15
EViews provides estimates for all three coefficients, @ , B 1. and Bo .
EViews provides estimates for all three coefficients, @ , B 1. and Bo .
ls log(volume) c t tsqr Estimating your first regression in EViews—17
ls log(volume) c t tsqr Estimating your first regression in EViews—17
Prior to this step the title bar of the equation window read “Equation: untitled.” Using th [Name] button changed two things: the equation now has a name which appears in the titl bar, and more importantly, the equation object is stored in the workfile. You can see thes: changes below. If you like, close the equation window and then double-click on [E] pre_break_equation to re-open the equation. But don’t take the break quite yet!
Prior to this step the title bar of the equation window read “Equation: untitled.” Using th [Name] button changed two things: the equation now has a name which appears in the titl bar, and more importantly, the equation object is stored in the workfile. You can see thes: changes below. If you like, close the equation window and then double-click on [E] pre_break_equation to re-open the equation. But don’t take the break quite yet!
What's Ahead This chapter’s been a quick stroll through EViews, just enough—we hope—to whet your appetite. You can continue walking through the chapters in order, but skipping around is fine too. If you’ll be mostly using EViews files prepared by others, you might proceed to Chapter 3, “Getting the Most from Least Squares,” to dive right into regressions; to
What's Ahead This chapter’s been a quick stroll through EViews, just enough—we hope—to whet your appetite. You can continue walking through the chapters in order, but skipping around is fine too. If you’ll be mostly using EViews files prepared by others, you might proceed to Chapter 3, “Getting the Most from Least Squares,” to dive right into regressions; to
Table 1: Academic Salary Data Excerpt
Table 1: Academic Salary Data Excerpt
Creating a New Workfile You'll notice in the dialog that EViews defaults to “Dated - regular frequency” and “Annual.” However, the data shown in Table 1: Academic Salary Data Excerpt are just num- bered sequentially. They aren’t dated.
Creating a New Workfile You'll notice in the dialog that EViews defaults to “Dated - regular frequency” and “Annual.” However, the data shown in Table 1: Academic Salary Data Excerpt are just num- bered sequentially. They aren’t dated.
These two methods create a series named NONACADSAL and place it in the workfile:
These two methods create a series named NONACADSAL and place it in the workfile:
We're ready to type numbers. But there’s a trick to entering your data. To protect against accidents, EViews locks the window so that it can’t be edited. To unlock, click on the button. Unlocked windows, as shown below for example, have an edit field just below the button bar. One way to know that a window is locked against editing is to observe the absence of the edit field. Alternatively, if you start typing and nothing happens, you’ll remember that you meant to click on the button—at least that’s what usually hap- pens to the author. Hint for the terminally obedient: For goodness sakes, don’t really enter all the data at the beginning of the chapter. You’ll be bored out of your mind. Just type in a few num- bers until you’re comfortable moving around in the window.
We're ready to type numbers. But there’s a trick to entering your data. To protect against accidents, EViews locks the window so that it can’t be edited. To unlock, click on the button. Unlocked windows, as shown below for example, have an edit field just below the button bar. One way to know that a window is locked against editing is to observe the absence of the edit field. Alternatively, if you start typing and nothing happens, you’ll remember that you meant to click on the button—at least that’s what usually hap- pens to the author. Hint for the terminally obedient: For goodness sakes, don’t really enter all the data at the beginning of the chapter. You’ll be bored out of your mind. Just type in a few num- bers until you’re comfortable moving around in the window.
Move up one cell from the default (you may want to use the scroll bar on the right hand side) and enter the word DISCIPLINE. EViews determines if you are specifying a series with numbers or with letters. Since DISCIPLINE is an empty alpha series, you will see empty val- ues in the DISCIPLINE column. Hint: EViews uses NA to indicate “not available” for numeric series and just an empty string for a “not available” alpha value. The latter explains why the observations for DISCIPLINE are blank.
Move up one cell from the default (you may want to use the scroll bar on the right hand side) and enter the word DISCIPLINE. EViews determines if you are specifying a series with numbers or with letters. Since DISCIPLINE is an empty alpha series, you will see empty val- ues in the DISCIPLINE column. Hint: EViews uses NA to indicate “not available” for numeric series and just an empty string for a “not available” alpha value. The latter explains why the observations for DISCIPLINE are blank.
When we set up an Unstructured/Undated workfile, EViews just numbers the observations 1, 2, 3, etc. (In dated workfiles, see below, dates are used for ids rather than sequence num- bers.) Rather than calling data for dentistry “observation 1” and data for medicine “observa- tion 2,” it might be a lot more meaningful to label them “dentistry” and “medicine.” EViews lets us specify that one of the existing series—obviously DISCIPLINE is the sensible choice— should be used as the id series.
When we set up an Unstructured/Undated workfile, EViews just numbers the observations 1, 2, 3, etc. (In dated workfiles, see below, dates are used for ids rather than sequence num- bers.) Rather than calling data for dentistry “observation 1” and data for medicine “observa- tion 2,” it might be a lot more meaningful to label them “dentistry” and “medicine.” EViews lets us specify that one of the existing series—obviously DISCIPLINE is the sensible choice— should be used as the id series.
Identity Noncrisis—33
Identity Noncrisis—33
Hint: Even when data are measured at regular intervals, measurements are sometimes missing. Not a problem, just leave missing measurements marked NA.
Hint: Even when data are measured at regular intervals, measurements are sometimes missing. Not a problem, just leave missing measurements marked NA.
Why add a date structure to a workfile? One minor reason is that it saves you the trouble of figuring out that 1947q1 through 2004q4 includes exactly 232 observations. There are two more important reasons:
Why add a date structure to a workfile? One minor reason is that it saves you the trouble of figuring out that 1947q1 through 2004q4 includes exactly 232 observations. There are two more important reasons:
Tips for dating: In addition to the Annual and Quarterly frequencies that we’ve seen, EViews offers a wide range of built-in dated regular frequencies for those of you with dated data: (Multi-year, Annual, Semi-annual, Quarterly, Monthly, Bimonthly, Fort- night, Ten-day, Weekly, Daily - 5 day week, Daily - 7 day week, Daily - custom week, Intraday), and a special frequency (Integer date) which is a generalization of Unstructured/Undated.
Tips for dating: In addition to the Annual and Quarterly frequencies that we’ve seen, EViews offers a wide range of built-in dated regular frequencies for those of you with dated data: (Multi-year, Annual, Semi-annual, Quarterly, Monthly, Bimonthly, Fort- night, Ten-day, Weekly, Daily - 5 day week, Daily - 7 day week, Daily - custom week, Intraday), and a special frequency (Integer date) which is a generalization of Unstructured/Undated.
Hint: Date functions work as expected in Dated - irregular workfiles. However, lags pick up the preceding observation (as in unstructured workfiles), not the preceding date (as in regular dated workfiles). In our original file, one lag of 11/27/1987 was 11/26/1987, which happened to be NA. In our new file, one lag of 11/27/1987 is 11/25/1987.
Hint: Date functions work as expected in Dated - irregular workfiles. However, lags pick up the preceding observation (as in unstructured workfiles), not the preceding date (as in regular dated workfiles). In our original file, one lag of 11/27/1987 was 11/26/1987, which happened to be NA. In our new file, one lag of 11/27/1987 is 11/25/1987.
Make It Slightly Easier Hint: Instead of choosing the menu File/Open/Foreign Data as Workfile... right click in any empty space inside EViews (not in the command pane or inside an open window) and choose Open/Foreign Data as Workifile.... Make It Really Easy Hint: Just drag-and-drop any data file onto an empty space inside EViews. If EViews understands the data in the file, the file will pop open, ready to read. (You may have to answer a couple of questions first.)
Make It Slightly Easier Hint: Instead of choosing the menu File/Open/Foreign Data as Workfile... right click in any empty space inside EViews (not in the command pane or inside an open window) and choose Open/Foreign Data as Workifile.... Make It Really Easy Hint: Just drag-and-drop any data file onto an empty space inside EViews. If EViews understands the data in the file, the file will pop open, ready to read. (You may have to answer a couple of questions first.)
The second-lowest common denomina- tor file format is a Microsoft Excel spreadsheet. Here’s an excerpt of “academic salaries by discipline.xls” (available on the EViews website). The second-lowest
The second-lowest common denomina- tor file format is a Microsoft Excel spreadsheet. Here’s an excerpt of “academic salaries by discipline.xls” (available on the EViews website). The second-lowest
YOU CaN CHCK In eacn column to change the series name or enter a description for the series. In our example EViews has correctly analyzed the file, so we can just hit and EViews generates our workfile.
YOU CaN CHCK In eacn column to change the series name or enter a description for the series. In our example EViews has correctly analyzed the file, so we can just hit and EViews generates our workfile.
The Import Business—41
The Import Business—41
The controls in the upper left hand corner of the second Spreadsheet read dialog provide a number of options for customizing how EViews interprets the Excel spreadsheet. Which option you need depends on how your file is structured. Here’s one example. The Header type: dropdown gives options that can handle most com- mon file arrangements. In this case, choose Names in last line and away we go. (Names in last line means the names are at the end of the header information, right before the data begins—a pretty common arrangement.)
The controls in the upper left hand corner of the second Spreadsheet read dialog provide a number of options for customizing how EViews interprets the Excel spreadsheet. Which option you need depends on how your file is structured. Here’s one example. The Header type: dropdown gives options that can handle most com- mon file arrangements. In this case, choose Names in last line and away we go. (Names in last line means the names are at the end of the header information, right before the data begins—a pretty common arrangement.)
It’s not at all unusual for a data file to include the date of each observation. EViews does a surprisingly good job of guessing that a particular column of data consists of dates that ought to be used as identifiers in the workfile. 0 Hint: “Text” data and files are variously described as “text,” “ASCII,” “alpha,” “alpha- numeric,” or “character.” Sometimes the file extension “csv” is used for text data where data values are separated by commas.
It’s not at all unusual for a data file to include the date of each observation. EViews does a surprisingly good job of guessing that a particular column of data consists of dates that ought to be used as identifiers in the workfile. 0 Hint: “Text” data and files are variously described as “text,” “ASCII,” “alpha,” “alpha- numeric,” or “character.” Sometimes the file extension “csv” is used for text data where data values are separated by commas.
The data in “academic salaries by discipline.txt” lines up pretty much the same way as did the same data in the Excel file we looked at above. To pull text data directly into an workfile use File/Open/Foreign Data as Workfile... and point to the appropriate text file. (You don’t want File/Open/Text File..., that’s for bringing a file in as text, not for converting the text to an EViews workfile.)
The data in “academic salaries by discipline.txt” lines up pretty much the same way as did the same data in the Excel file we looked at above. To pull text data directly into an workfile use File/Open/Foreign Data as Workfile... and point to the appropriate text file. (You don’t want File/Open/Text File..., that’s for bringing a file in as text, not for converting the text to an EViews workfile.)
Hint: Telling EViews to use fixed column locations for each series replaces the use of delimiters. This can get around the missing number problem described above. Simi- larly, it can solve the problem of alpha observations that include spaces that would be mistaken for delimiters. The Import Business—47
Hint: Telling EViews to use fixed column locations for each series replaces the use of delimiters. This can get around the missing number problem described above. Simi- larly, it can solve the problem of alpha observations that include spaces that would be mistaken for delimiters. The Import Business—47
Hint: Sometimes the Open dialog remains open for what seems like a long time whil EViews processes the data from the web. Be patient.
Hint: Sometimes the Open dialog remains open for what seems like a long time whil EViews processes the data from the web. Be patient.
Copy/Paste
Copy/Paste
We want to extend the range of the first workfile and then copy in the data from the second. Select the workfile window for “First Two.wfl”. Use File/SaveAs to change the name to “All Data”. Now double-click on Range and change the range to 5 observations. Adding Data To An Existing Workfile—Or, Being Rectangular Doesn't Mean Being Inflexible—53
We want to extend the range of the first workfile and then copy in the data from the second. Select the workfile window for “First Two.wfl”. Use File/SaveAs to change the name to “All Data”. Now double-click on Range and change the range to 5 observations. Adding Data To An Existing Workfile—Or, Being Rectangular Doesn't Mean Being Inflexible—53
We need to be careful which workfile we’re working in now. Select the workfile window for “Last Three.wfl”. Hit the button and Select All (except C-Resid). Then open a group window, which will look more or less like the second window above. Select all the data by dragging the mouse. Then copy to the clipboard.
We need to be careful which workfile we’re working in now. Select the workfile window for “Last Three.wfl”. Hit the button and Select All (except C-Resid). Then open a group window, which will look more or less like the second window above. Select all the data by dragging the mouse. Then copy to the clipboard.
Hint: To prevent scrambling up the variables, be sure to move data for all your series at the same time. Adding Data To An Existing Workfile—Or, Being Rectangular Doesn’t Mean Being Inflexible—55
Hint: To prevent scrambling up the variables, be sure to move data for all your series at the same time. Adding Data To An Existing Workfile—Or, Being Rectangular Doesn’t Mean Being Inflexible—55
Handling multiple missing codes If you prefer, you can accomplish the same task with the recode command, as in:
Handling multiple missing codes If you prefer, you can accomplish the same task with the recode command, as in:
Hint: Since dates are stored as numbers, you can do sensible date arithmetic. If the series DATEBOUGHT and DATESOLD hold the information suggested by their respec- tive names, then: Appendix: Having A Good Time With Your Date—59 So dates are pretty straightforward. Except when they’re not. If you want more details, the Command and Programming Reference has a very nice 20+ page section for you.
Hint: Since dates are stored as numbers, you can do sensible date arithmetic. If the series DATEBOUGHT and DATESOLD hold the information suggested by their respec- tive names, then: Appendix: Having A Good Time With Your Date—59 So dates are pretty straightforward. Except when they’re not. If you want more details, the Command and Programming Reference has a very nice 20+ page section for you.
A First Regression
A First Regression
Hint: EViews tells one item in a list from another by looking for spaces between items. For this reason, spaces generally aren’t allowed inside a single item. If you type:
Hint: EViews tells one item in a list from another by looking for spaces between items. For this reason, spaces generally aren’t allowed inside a single item. If you type:
The Really Important Regression Results There are 25 pieces of information displayed for this very simple regression. To sort out all the different goodies, we’ll start by showing a couple of ways that the main results might be presented in a scientific paper. Then we’ll discuss the remaining items one number at a time. Hint: Another reminder: @t rend is an EViews function to generate a time trend, 0, 1, Poss nen
The Really Important Regression Results There are 25 pieces of information displayed for this very simple regression. To sort out all the different goodies, we’ll start by showing a couple of ways that the main results might be presented in a scientific paper. Then we’ll discuss the remaining items one number at a time. Hint: Another reminder: @t rend is an EViews function to generate a time trend, 0, 1, Poss nen
Column (2)? Don’t worry, we’ll come back to it later. The convention for inline reporting works well for a single equation, but becomes unwieldy when you have more than one equation to report. Results from several related regressions might be displayed in a table, looking something like Table 2. Hint: Good scientific practice is to report only digits that are meaningful when display- ing a number. We’ve printed far too many digits in both the inline display and in Table 2 so as to make it easy for you to match up the displayed numbers with the EViews output. From now on we’ll be better behaved.
Column (2)? Don’t worry, we’ll come back to it later. The convention for inline reporting works well for a single equation, but becomes unwieldy when you have more than one equation to report. Results from several related regressions might be displayed in a table, looking something like Table 2. Hint: Good scientific practice is to report only digits that are meaningful when display- ing a number. We’ve printed far too many digits in both the inline display and in Table 2 so as to make it easy for you to match up the displayed numbers with the EViews output. From now on we’ll be better behaved.
t-Tests and Stuff The column headed “t-Statistic” reports, not surprisingly, the t-statistic. Specifically, this is the t-statistic for the hypothesis that the coefficient in the same row equals zero. (It’s com- puted as the ratio of the estimated coefficient to its standard error: e.g., 51.7 = 0.017/0.00033 .)
t-Tests and Stuff The column headed “t-Statistic” reports, not surprisingly, the t-statistic. Specifically, this is the t-statistic for the hypothesis that the coefficient in the same row equals zero. (It’s com- puted as the ratio of the estimated coefficient to its standard error: e.g., 51.7 = 0.017/0.00033 .)
Summary Regression Statistics We’ve already talked about the two most important numbers, “R-squared” and “S.E. of regression.” Our regression accounts for 85 percent of the variance in the dependent vari- able and the estimated standard deviation of the error term is 0.97. Five other elements, “Sum squared residuals,” “Log likelihood,” “Akaike info criterion,” “Schwarz criterion,” and “Hannan-Quinn criter.” are used for making statistical comparisons between two different regressions. This means that they don’t really help us learn anything about the regression we're working on; rather, these statistics are useful for deciding if one model is better than another. For the record, the sum of squared residuals is used in computing F-tests, the log likelihood is used for computing likelihood ratio tests, and the Akaike and Schwarz criteria are used in Bayesian model comparison.
Summary Regression Statistics We’ve already talked about the two most important numbers, “R-squared” and “S.E. of regression.” Our regression accounts for 85 percent of the variance in the dependent vari- able and the estimated standard deviation of the error term is 0.97. Five other elements, “Sum squared residuals,” “Log likelihood,” “Akaike info criterion,” “Schwarz criterion,” and “Hannan-Quinn criter.” are used for making statistical comparisons between two different regressions. This means that they don’t really help us learn anything about the regression we're working on; rather, these statistics are useful for deciding if one model is better than another. For the record, the sum of squared residuals is used in computing F-tests, the log likelihood is used for computing likelihood ratio tests, and the Akaike and Schwarz criteria are used in Bayesian model comparison.
Hint: Most regression specifications include an intercept. Be sure to include “C” in the list of independent variables unless you’re sure you don’t want an intercept. ls log(volume) c @trend @trend*%2 log (volume (-1)
Hint: Most regression specifications include an intercept. Be sure to include “C” in the list of independent variables unless you’re sure you don’t want an intercept. ls log(volume) c @trend @trend*%2 log (volume (-1)
Hint: The p-value reported by EViews is computed for a two-tailed test. If you’re inter- ested in a one-tailed test, you’ll have to look up the critical value for yourself. 76—Chapter 3. Getting the Most from Least Squares Suppose we wanted to test whether the coefficient on LOG(VOLUME(-1)) equaled one rather than zero. Enter “c(4) =1” to find the new test statistic. EJ Equation: UNTITLED Workfile: NYSEVOLUME::.. — ©
Hint: The p-value reported by EViews is computed for a two-tailed test. If you’re inter- ested in a one-tailed test, you’ll have to look up the critical value for yourself. 76—Chapter 3. Getting the Most from Least Squares Suppose we wanted to test whether the coefficient on LOG(VOLUME(-1)) equaled one rather than zero. Enter “c(4) =1” to find the new test statistic. EJ Equation: UNTITLED Workfile: NYSEVOLUME::.. — ©
Econometric theory warning: If you’ve studied the advanced topic in econometric the- ory called the “unit root problem” you know that standard theory doesn’t apply in this test (although the issue is harmless for this particular set of data). Take this as a reminder that you and EViews are a team, but you’re the brains of the outfit. EViews will obediently do as it’s told. It’s up to you to choose the proper procedure.
Econometric theory warning: If you’ve studied the advanced topic in econometric the- ory called the “unit root problem” you know that standard theory doesn’t apply in this test (although the issue is harmless for this particular set of data). Take this as a reminder that you and EViews are a team, but you’re the brains of the outfit. EViews will obediently do as it’s told. It’s up to you to choose the proper procedure.
78—Chapter 3. Getting the Most from Least Squares
78—Chapter 3. Getting the Most from Least Squares
Hint: Okay, okay. Maybe you didn’t really need the representations view as a reminder. The real value of this view is that you can copy the equation from this view and then paste it into your word processor, or into an EViews batch program, or even into Excel, where with a little judicious editing you can turn the equation into an Excel formula.
Hint: Okay, okay. Maybe you didn’t really need the representations view as a reminder. The real value of this view is that you can copy the equation from this view and then paste it into your word processor, or into an EViews batch program, or even into Excel, where with a little judicious editing you can turn the equation into an Excel formula.
What's Left After You've Gotten the Most Out of Least Squares—79
What's Left After You've Gotten the Most Out of Least Squares—79
Your Basic Elementary Algebra
Your Basic Elementary Algebra
Deconstructing Two Seconds Construction
Deconstructing Two Seconds Construction
Some Typing Issues Hint: EViews doesn’t care about the capitalization of commands or series names. If you’ve been busy entering a lot of commands, you may press (CTRL+J) to examine a his- tory of the last 30 commands. Use the UP and DOWN arrows to select the desired command and press ENTER, or double click on the desired command to add it to the command win- dow. To close the history window without selecting a command, click elsewhere in the command window or press the Escape (ESC) key.
Some Typing Issues Hint: EViews doesn’t care about the capitalization of commands or series names. If you’ve been busy entering a lot of commands, you may press (CTRL+J) to examine a his- tory of the last 30 commands. Use the UP and DOWN arrows to select the desired command and press ENTER, or double click on the desired command to add it to the command win- dow. To close the history window without selecting a command, click elsewhere in the command window or press the Escape (ESC) key.
You can print the command pane by clicking anywhere in the pane and then choosing File/Print. Similarly, you can save the command pane to disk (default file name “com- mand.log”) by clicking anywhere in the pane and choosing File/Save or File/SaveAs.... Deprecatory hint: Earlier versions of EViews used the command genr for what’s now done with the distinct commands series, alpha, and frml. (We'll meet the latter two commands shortly.) Genr will still work even though the new commands are pre- ferred. (Computer folks say an old feature has been “deprecated” when it’s been replaced by something new, but the old feature continues to work.)
You can print the command pane by clicking anywhere in the pane and then choosing File/Print. Similarly, you can save the command pane to disk (default file name “com- mand.log”) by clicking anywhere in the pane and choosing File/Save or File/SaveAs.... Deprecatory hint: Earlier versions of EViews used the command genr for what’s now done with the distinct commands series, alpha, and frml. (We'll meet the latter two commands shortly.) Genr will still work even though the new commands are pre- ferred. (Computer folks say an old feature has been “deprecated” when it’s been replaced by something new, but the old feature continues to work.)
show january @weekday=5
show january @weekday=5
Your Basic Elementary Algebra—93 Then we can run the usual regression with:
Your Basic Elementary Algebra—93 Then we can run the usual regression with:
Typographic hint: EViews thinks series are separated by spaces. This means that when using an auto-series it’s important that there not be any spaces, unless the auto-series is enclosed in parentheses. In our example, EViews interprets y-@mean (y) as the deviation of Y from its mean, as we intended. If we had left a space before the minus sign, EViews would have thought we wanted Y to be the dependent variable and the first independent variable to be the negative of the mean of Y.
Typographic hint: EViews thinks series are separated by spaces. This means that when using an auto-series it’s important that there not be any spaces, unless the auto-series is enclosed in parentheses. In our example, EViews interprets y-@mean (y) as the deviation of Y from its mean, as we intended. If we had left a space before the minus sign, EViews would have thought we wanted Y to be the dependent variable and the first independent variable to be the negative of the mean of Y.
A sample is an EViews object which specifies which observations are to be included in oper- ations. Effectively, you can think of a sample as a list of dates in a dated workfile, or a list of observation numbers in an undated workfile. Samples are used for operational control in two different places. The primary sample is the workfile sample. This sample provides the default control for all operations on series, by telling which observations to include. Specific commands occasionally allow specification of a secondary sample which over-rides the workfile sample.
A sample is an EViews object which specifies which observations are to be included in oper- ations. Effectively, you can think of a sample as a list of dates in a dated workfile, or a list of observation numbers in an undated workfile. Samples are used for operational control in two different places. The primary sample is the workfile sample. This sample provides the default control for all operations on series, by telling which observations to include. Specific commands occasionally allow specification of a secondary sample which over-rides the workfile sample.
Three special keywords help out in specifying date pairs. @first means the first date in the workfile. @last means the last date. @al1 means all dates in the workfile. So two equiva- lent commands are: smpl 1/03/2005 1/03/2005 1/05/2005 1/07/2005
Three special keywords help out in specifying date pairs. @first means the first date in the workfile. @last means the last date. @al1 means all dates in the workfile. So two equiva- lent commands are: smpl 1/03/2005 1/03/2005 1/05/2005 1/07/2005
Significant digits Fixed decimal Scientific notation Percentage Fraction Period Day Day-Time Time Number Display Numbers and Letters—101
Significant digits Fixed decimal Scientific notation Percentage Fraction Period Day Day-Time Time Number Display Numbers and Letters—101
EViews’ second major data type is text data, stored in an alpha series displayed with the icon. You create an alpha series by typing the command:
EViews’ second major data type is text data, stored in an alpha series displayed with the icon. You create an alpha series by typing the command:
General Options/Series and Alphas/Alpha truncation dialog. And you should. Again, see Chapter 18, “Optional Ending.” Alpha series have two quirks that matter once in a great while. The first quirk is that the “not available” code for alpha series is the empty string, “”, rather than NA. Visually, it’s difficult to tell the difference between an empty string and a string holding nothing but one or more spaces. They both have a blank look about them.
General Options/Series and Alphas/Alpha truncation dialog. And you should. Again, see Chapter 18, “Optional Ending.” Alpha series have two quirks that matter once in a great while. The first quirk is that the “not available” code for alpha series is the empty string, “”, rather than NA. Visually, it’s difficult to tell the difference between an empty string and a string holding nothing but one or more spaces. They both have a blank look about them.
Numbers and Letters—105
Numbers and Letters—105
gives a numerical date series which is then available for further manipulation. The function @datestr(x[,fmt]) translates the number in X into a text string. A second optional argument specifies a format for the date. As examples, @datestr (731946) pro- duces “1/1/2005”; @datestr (731946,"Month dd, yyyy") gives “January 1, 2005”; and @datestr (731946, "ww") produces a text string representation of the week number, “6.” More date formats are discussed in the User’s Guide. @makedate translates ordinary numbers into date numbers, for example @makedate (1999, "yyyy") returns 729754, the first day in 1999. The format strings usec for the last argument of @makedate are also discussed in the User’s Guide.
gives a numerical date series which is then available for further manipulation. The function @datestr(x[,fmt]) translates the number in X into a text string. A second optional argument specifies a format for the date. As examples, @datestr (731946) pro- duces “1/1/2005”; @datestr (731946,"Month dd, yyyy") gives “January 1, 2005”; and @datestr (731946, "ww") produces a text string representation of the week number, “6.” More date formats are discussed in the User’s Guide. @makedate translates ordinary numbers into date numbers, for example @makedate (1999, "yyyy") returns 729754, the first day in 1999. The format strings usec for the last argument of @makedate are also discussed in the User’s Guide.
vations, giving us 365/1. Over an ordinary weekend, the datediff function returns 3, sc we annualize by multiplying by 365/3.
vations, giving us 365/1. Over an ordinary weekend, the datediff function returns 3, sc we annualize by multiplying by 365/3.
Equation: UNTITLED Workfile: CPSMAR2004EXTRACT::CPSt Hint: Labels in the editor are case sensitive, e.g., “male” translates to 0 but “Male” translates to NA. What Are Your Values?—111
Equation: UNTITLED Workfile: CPSMAR2004EXTRACT::CPSt Hint: Labels in the editor are case sensitive, e.g., “male” translates to 0 but “Male” translates to NA. What Are Your Values?—111
EViews has lots of functions for transforming data. You’ll never need most of these func- tions, but the one you do need you’ll need bad. Value maps can be used to group a range of codes for the purpose of display. Instead of a sin- gle value in the value map, enter a range in parentheses. For example “(-inf, 12)” specifies all values less than 12. Parentheses are used to specify open intervals, square brackets are used for closed intervals. So “(-inf, 12]” is all values less than or equal to 12.
EViews has lots of functions for transforming data. You’ll never need most of these func- tions, but the one you do need you’ll need bad. Value maps can be used to group a range of codes for the purpose of display. Instead of a sin- gle value in the value map, enter a range in parentheses. For example “(-inf, 12)” specifies all values less than 12. Parentheses are used to specify open intervals, square brackets are used for closed intervals. So “(-inf, 12]” is all values less than or equal to 12.
show f d @meansby(ed,fe) @stdevsby (ed, fe) Expand the Dummies
show f d @meansby(ed,fe) @stdevsby (ed, fe) Expand the Dummies
tions.) These functions summarize the data in X according to the groups in Y. (Optionally, a sample can be used as a third argument.) Thus the command:
tions.) These functions summarize the data in X according to the groups in Y. (Optionally, a sample can be used as a third argument.) Thus the command:
A variety of probability distributions are discussed in the Command and Programming Refer- ence. Probably the most commonly used are “norm,” for standard normal, and “tdist,” for Student’s ¢. Here are a few examples:
A variety of probability distributions are discussed in the Command and Programming Refer- ence. Probably the most commonly used are “norm,” for standard normal, and “tdist,” for Student’s ¢. Here are a few examples:
Interest rates over a wide spectrum of maturities—three months to 20 years—mostly move up and down together. Long term interest rates are usually, although not always, higher than short term interest rates. Long term interest rates also bounce around less than short term interest rates. One picture illustrates all this at a glance.
Interest rates over a wide spectrum of maturities—three months to 20 years—mostly move up and down together. Long term interest rates are usually, although not always, higher than short term interest rates. Long term interest rates also bounce around less than short term interest rates. One picture illustrates all this at a glance.
Get A Closer Look To Freeze Or Not To Freeze Before adding any customizations, we have a choice to make about whether to freeze the graph. The group we’re looking at right now is fundamentally a list of data series, which we happen to be looking at in a graphics view. If we change any of the underlying data, the change will be reflected in the picture. Same thing if we change the sample. For that matter, we can switch to a different graphics view or even a spreadsheet or statistical view. But a group view does have one shortcoming:
Get A Closer Look To Freeze Or Not To Freeze Before adding any customizations, we have a choice to make about whether to freeze the graph. The group we’re looking at right now is fundamentally a list of data series, which we happen to be looking at in a graphics view. If we change any of the underlying data, the change will be reflected in the picture. Same thing if we change the sample. For that matter, we can switch to a different graphics view or even a spreadsheet or statistical view. But a group view does have one shortcoming:
Hint: If you select Off and then decide you’d like to relate the graph object to its under- lying data again, you can always change your selection later in the master Graph Options dialog, in the Graph Updating section. A graph object is fundamentally a picture that happens to have started life as a graphic view. You make a graph object by looking at a graphic view, as we are at the moment, and hitting the button.
Hint: If you select Off and then decide you’d like to relate the graph object to its under- lying data again, you can always change your selection later in the master Graph Options dialog, in the Graph Updating section. A graph object is fundamentally a picture that happens to have started life as a graphic view. You make a graph object by looking at a graphic view, as we are at the moment, and hitting the button.
Hint: Because a frozen graph with updating off is severed from the underlying data series, the options for changing from one type of graph to another (categorical graph, distribution plot, etc.) are limited. It’s generally best to choose a graph type before freezing a graph if you intend to keep updating off. A Simple Soup-To-Nuts Graphing Example—121
Hint: Because a frozen graph with updating off is severed from the underlying data series, the options for changing from one type of graph to another (categorical graph, distribution plot, etc.) are limited. It’s generally best to choose a graph type before freezing a graph if you intend to keep updating off. A Simple Soup-To-Nuts Graphing Example—121
A Little Light Customization Whenever more than one series appears on a graph, the question arises as to how to visu- ally distinguish one graphed line from another. The two methods are to use different colors and to use different line patterns. Different colors are much easier to distinguish—unless your output device only shows black and white!
A Little Light Customization Whenever more than one series appears on a graph, the question arises as to how to visu- ally distinguish one graphed line from another. The two methods are to use different colors and to use different line patterns. Different colors are much easier to distinguish—unless your output device only shows black and white!
default, the graph appears in solid lines distinguished by colors on your display screen, bu patterned black lines are used if you print from EViews to a monochrome printer. A Simple Soup-To-Nuts Graphing Example—123
default, the graph appears in solid lines distinguished by colors on your display screen, bu patterned black lines are used if you print from EViews to a monochrome printer. A Simple Soup-To-Nuts Graphing Example—123
In making an aesthetically pleasing data graph, EViews hides the details of many complex calculations. Graphic output is tuned with many small tweaks to make the graph look “just right.” In particular, well done graphics scale nonlinearly. In other words, if you double the picture size, picture elements don’t simply grow to twice their original size. If you plan to print or publish a graphic, try to make it as close as possible to its final appearance while it’s still inside EViews. Pay particular attention to color versus monochrome and to the ratio of height versus width. Print, Copy, Or Export
In making an aesthetically pleasing data graph, EViews hides the details of many complex calculations. Graphic output is tuned with many small tweaks to make the graph look “just right.” In particular, well done graphics scale nonlinearly. In other words, if you double the picture size, picture elements don’t simply grow to twice their original size. If you plan to print or publish a graphic, try to make it as close as possible to its final appearance while it’s still inside EViews. Pay particular attention to color versus monochrome and to the ratio of height versus width. Print, Copy, Or Export
Hint: There isn’t any way to read a graphic file into EViews, nor can you paste a pic ture from the clipboard into an EViews object.
Hint: There isn’t any way to read a graphic file into EViews, nor can you paste a pic ture from the clipboard into an EViews object.
Or suppose we wanted a scatter plot of long rates against the short (3-month) rate? Just choose the Scatter and accept the defaults settings. 128—Chapter 5. Picture This!
Or suppose we wanted a scatter plot of long rates against the short (3-month) rate? Just choose the Scatter and accept the defaults settings. 128—Chapter 5. Picture This!
But here’s my favorite one-click wonder: Change the Graph type back to Line & Symbol and then with a single click, change Graph data: to Means: A Graphic Description of the Creative Process—129
But here’s my favorite one-click wonder: Change the Graph type back to Line & Symbol and then with a single click, change Graph data: to Means: A Graphic Description of the Creative Process—129
Financial econometrics visualization alert: This average yield curve worked out neatly because the series in the group “just happened to be” ordered from short maturity to long maturity. If we’d chosen a different order for the series, the line connecting the means wouldn’t have been meaningful. As is, the scaling on the horizontal axis is a lit- tle misleading. We probably think of the one-year rate as being close to the 3-month rate, not halfway between the 3-month and 20-year rates.
Financial econometrics visualization alert: This average yield curve worked out neatly because the series in the group “just happened to be” ordered from short maturity to long maturity. If we’d chosen a different order for the series, the line connecting the means wouldn’t have been meaningful. As is, the scaling on the horizontal axis is a lit- tle misleading. We probably think of the one-year rate as being close to the 3-month rate, not halfway between the 3-month and 20-year rates.
The standard line graph to the right shows U.S. retail and food service sales over a dozen years. Notice the regular spikes. How fas can you say “Christmas?”
The standard line graph to the right shows U.S. retail and food service sales over a dozen years. Notice the regular spikes. How fas can you say “Christmas?”
Picture One Series—135
Picture One Series—135
Distribution, Quantile-Quantile, and Boxplots
Distribution, Quantile-Quantile, and Boxplots
Axis Borders
Axis Borders
Each of the graphs in the window is a separate graphic sub-object. You can set options for the graphs individually or all together. You can also grab the graphs with the mouse and drag them around to re-arrange their locations in the window. 138—Chapter 5. Picture This!
Each of the graphs in the window is a separate graphic sub-object. You can set options for the graphs individually or all together. You can also grab the graphs with the mouse and drag them around to re-arrange their locations in the window. 138—Chapter 5. Picture This!
Hint: If you prefer, EViews can auto-arrange the individual graphs into neat rows and columns. Right-click on the graph window and choose Position and align graphs...
Hint: If you prefer, EViews can auto-arrange the individual graphs into neat rows and columns. Right-click on the graph window and choose Position and align graphs...
Jne more option is immediately relevant. Return to the
Jne more option is immediately relevant. Return to the
ihe exampie here shows high and low prices for the Russell 3000 (a very broad index of U.S. stocks) in the later part of 1987.
ihe exampie here shows high and low prices for the Russell 3000 (a very broad index of U.S. stocks) in the later part of 1987.
Mixed With Lines High-Low — High-Low-Open-Close Graphs
Mixed With Lines High-Low — High-Low-Open-Close Graphs
Hint: The legend isn’t displayed on High-Low graphs, so be sure to include your own label using the button as we have here.
Hint: The legend isn’t displayed on High-Low graphs, so be sure to include your own label using the button as we have here.
Orderly hint: The order in which the series appear in the group matters for the High- Low graph. EViews puts series in the same order as you select them in the workfile window, but the order is easily re-arranged by choosing the Group Members view and manually editing the list. This is true for any group, but it’s especially important for graphs such as the High-Low graph, where the order of the series affects their inter- pretation.
Orderly hint: The order in which the series appear in the group matters for the High- Low graph. EViews puts series in the same order as you select them in the workfile window, but the order is easily re-arranged by choosing the Group Members view and manually editing the list. This is true for any group, but it’s especially important for graphs such as the High-Low graph, where the order of the series affects their inter- pretation.
show gdpc96f+1.96*gdpc9b6se gdpc9Ib6f-1.96*gdpc9bse gdpc9b6Ff
show gdpc96f+1.96*gdpc9b6se gdpc9Ib6f-1.96*gdpc9bse gdpc9b6Ff
Scatter Plots Scatter plots are used for looking at the relation between two—or more—variables. We’ll use data on undergraduate grades and LSAT scores for applicants to the University of Washing- ton law school to illustrate scatter plots (“Law 98.wf1”).
Scatter Plots Scatter plots are used for looking at the relation between two—or more—variables. We’ll use data on undergraduate grades and LSAT scores for applicants to the University of Washing- ton law school to illustrate scatter plots (“Law 98.wf1”).
Simple Scatter Hint: Scatter plots don’t give good visuals when there are too many observations. The plots tend to get too “busy” to discern patterns. Additionally, scatter plots with very large numbers of observations can take a very, very long time to render when the graphs are copied outside of EViews.
Simple Scatter Hint: Scatter plots don’t give good visuals when there are too many observations. The plots tend to get too “busy” to discern patterns. Additionally, scatter plots with very large numbers of observations can take a very, very long time to render when the graphs are copied outside of EViews.
Hint: The equation for the line shown is, of course, the equation estimated by the com- mand
Hint: The equation for the line shown is, of course, the equation estimated by the com- mand
Scatter with Regression
Scatter with Regression
Options dialog. The default is to add more scatterings to the
Options dialog. The default is to add more scatterings to the
Togetherness of the First Sort
Togetherness of the First Sort
Using the right specification sure fits the data a lot better! Clever observation: Did you notice that EViews cleverly solved for Y even though we specified log(Y) on the left of the series command?
Using the right specification sure fits the data a lot better! Clever observation: Did you notice that EViews cleverly solved for Y even though we specified log(Y) on the left of the series command?
XY Line and XY Area Graphs The series INF_EARLY has inflation from 1959 through 1969 and NAs elsewhere. Similarly, INF_LATE is NA except for 1970 through 1979. Now we open a group with the unemploy- ment rate as the first series and INF_EARLY and INF_LATE as the second and third series. We can exploit the fact that EViews doesn’t plot observations with NA values to get a very nice looking XY Line graph.
XY Line and XY Area Graphs The series INF_EARLY has inflation from 1959 through 1969 and NAs elsewhere. Similarly, INF_LATE is NA except for 1970 through 1979. Now we open a group with the unemploy- ment rate as the first series and INF_EARLY and INF_LATE as the second and third series. We can exploit the fact that EViews doesn’t plot observations with NA values to get a very nice looking XY Line graph.
Obscure graph type hint: The graph type XY Bar (X-X-Y triplets) lets you draw bar graphs semi-manually. Vertical bars are drawn with the left edge specified by the first series, the right edge specified by the second series, and the height given by the third series.
Obscure graph type hint: The graph type XY Bar (X-X-Y triplets) lets you draw bar graphs semi-manually. Vertical bars are drawn with the left edge specified by the first series, the right edge specified by the second series, and the height given by the third series.
Hint: In case you were wondering, there’s no way to automatically label individual slices.
Hint: In case you were wondering, there’s no way to automatically label individual slices.
To twist a graph on it’s side, choose Rotated - obs axis on left in the Orientation combo of the Graph Type dialog. Below is a rotated ver- sion of the bar graph we saw on page 139.
To twist a graph on it’s side, choose Rotated - obs axis on left in the Orientation combo of the Graph Type dialog. Below is a rotated ver- sion of the bar graph we saw on page 139.
Let’s Look At This From Another Angle—155
Let’s Look At This From Another Angle—155
Pretty boring, eh? Even if you’re fascinated by wage distributions, that’s a pretty boring graph. All the choices other than Raw data produce a single summary statistic for each group of data. If all you have is one series, there’s only one number to plot. Plotting sum- mary statistics gets interesting when you compare statistics for different groups of data. We saw this in the comparison of the means of three different interest rate series in the plot of the average yield curve on page 130. We’ll see examples where the groups of data represent different categories in the next section. To visually summarize your data, change the Graph data: dropdown in the Details: field to a summary statistic of your choice. For example, here’s a bar graph showing the median level of wages and salaries for U.S. workers in 2004 (“cpsmar2004extract.wfl”).
Pretty boring, eh? Even if you’re fascinated by wage distributions, that’s a pretty boring graph. All the choices other than Raw data produce a single summary statistic for each group of data. If all you have is one series, there’s only one number to plot. Plotting sum- mary statistics gets interesting when you compare statistics for different groups of data. We saw this in the comparison of the means of three different interest rate series in the plot of the average yield curve on page 130. We’ll see examples where the groups of data represent different categories in the next section. To visually summarize your data, change the Graph data: dropdown in the Details: field to a summary statistic of your choice. For example, here’s a bar graph showing the median level of wages and salaries for U.S. workers in 2004 (“cpsmar2004extract.wfl”).
So far, all our graphs have produced one plot per series. EViews can also display plots of series broken down by one or more categories. Categorical graph This is a great tool for getting an idea of how one variable affects another. Categorical graphs work for both raw data and summary statistics and pretty mu all the graph types available under Basic graph are also available under Categorical gray
So far, all our graphs have produced one plot per series. EViews can also display plots of series broken down by one or more categories. Categorical graph This is a great tool for getting an idea of how one variable affects another. Categorical graphs work for both raw data and summary statistics and pretty mu all the graph types available under Basic graph are also available under Categorical gray
Splitting up wages by gender as well as union membership, we have four plots with numer- ous arrangement possibilities. If we make “UNION FE” the within factors (FE codes gender), we get the plot shown below.
Splitting up wages by gender as well as union membership, we have four plots with numer- ous arrangement possibilities. If we make “UNION FE” the within factors (FE codes gender), we get the plot shown below.
Note that the first categorical variable listed becomes the major grouping control. Switch the within order to “FE UNION” and EViews switches the ordering in the graph as shown below.
Note that the first categorical variable listed becomes the major grouping control. Switch the within order to “FE UNION” and EViews switches the ordering in the graph as shown below.
Hint: The graphs give identical information, but the first graph gives visual emphasis to the fact that men are paid more than women whether they’re union or non-union. The second graph emphasized the union wage premium for both genders.
Hint: The graphs give identical information, but the first graph gives visual emphasis to the fact that men are paid more than women whether they’re union or non-union. The second graph emphasized the union wage premium for both genders.
You can use as many within and across factors as you wish. We’ve added “highest degree received” to the across factor in this graph. If we move the variable FE to the Across field, EViews splits the graph across separate plots for each gender.
You can use as many within and across factors as you wish. We’ve added “highest degree received” to the across factor in this graph. If we move the variable FE to the Across field, EViews splits the graph across separate plots for each gender.
If we’d set Treat multiple series in this Group object as: First within factor both WSAL_VAL and HRSWK would appear in the same plot, as shown below. Because of the di ference in scales for the two series, First within factor wouldn’t be a sensible choice for thi application. series in this Group object as: menu to treat the series (the series in this group were WSAL_VAL and HRSWK) as a factor.
If we’d set Treat multiple series in this Group object as: First within factor both WSAL_VAL and HRSWK would appear in the same plot, as shown below. Because of the di ference in scales for the two series, First within factor wouldn’t be a sensible choice for thi application. series in this Group object as: menu to treat the series (the series in this group were WSAL_VAL and HRSWK) as a factor.
Polishing Factor Layouts 162—Chapter 5. Picture This!
Polishing Factor Layouts 162—Chapter 5. Picture This!
Hint: Categorical options aren’t available on frozen graphs if updating is disabled. Distinguishing Factors In the graph depicted earlier, all the bars are a single color and pattern. Contrast the graph to the graph shown below, where “union” bars are shown in solid red and “nonunion bars” come in cross-hatched blue. Turning on Within graph category identification instructs EViews to add visual distinction to the within graph elements. We’ve done this to change
Hint: Categorical options aren’t available on frozen graphs if updating is disabled. Distinguishing Factors In the graph depicted earlier, all the bars are a single color and pattern. Contrast the graph to the graph shown below, where “union” bars are shown in solid red and “nonunion bars” come in cross-hatched blue. Turning on Within graph category identification instructs EViews to add visual distinction to the within graph elements. We’ve done this to change
EViews interprets “add visual distinction” for this graph as assigning a unique color for each within category. This choice is ideal when the graph is presented in color, but we wanted clear visual distinction for a monochrome version as well. So we added the cross-hatching manually. To see how, see Fill Areas in the next chapter.
EViews interprets “add visual distinction” for this graph as assigning a unique color for each within category. This choice is ideal when the graph is presented in color, but we wanted clear visual distinction for a monochrome version as well. So we added the cross-hatching manually. To see how, see Fill Areas in the next chapter.
The second item worth knowing about this dialog is that it uses @series as a special key- word. You can see an example in the dialog on page 162. When a graph has multiple series, EViews will treat data coming from one series versus another analogously with data within a series coming from one category versus another. In other words, the list of series can be treated like an artificial categorical variable for arranging graph layouts. (Just as we saw in Multiple Series as Factors.) The keyword @series is used to identify the list of series in the graph.
The second item worth knowing about this dialog is that it uses @series as a special key- word. You can see an example in the dialog on page 162. When a graph has multiple series, EViews will treat data coming from one series versus another analogously with data within a series coming from one category versus another. In other words, the list of series can be treated like an artificial categorical variable for arranging graph layouts. (Just as we saw in Multiple Series as Factors.) The keyword @series is used to identify the list of series in the graph.
Using the mouse, you can re-arrange the position of the subgraphs within the overall graph window. This allows you to produce some very interesting effects. For example, if you have two graphs with identical axes, you can superimpose one on the other.
Using the mouse, you can re-arrange the position of the subgraphs within the overall graph window. This allows you to produce some very interesting effects. For example, if you have two graphs with identical axes, you can superimpose one on the other.
where tm3_ty01 tm3_ty20 are names of graphs stored in the workfile, to open a new graph object.
where tm3_ty01 tm3_ty20 are names of graphs stored in the workfile, to open a new graph object.
EViews does a masterful job of creating aesthetically pleasing graphs. But sometimes you want to tweak the picture to get it “just so,” choosing custom labeling, axes, colors, etc. In this chapter we get close-up and intimate with EViews graphics. The menu provides an editing feature that isn’t replicated with a button: Draw arrow. Chapter 6. Intimacy With Graphic Objects
EViews does a masterful job of creating aesthetically pleasing graphs. But sometimes you want to tweak the picture to get it “just so,” choosing custom labeling, axes, colors, etc. In this chapter we get close-up and intimate with EViews graphics. The menu provides an editing feature that isn’t replicated with a button: Draw arrow. Chapter 6. Intimacy With Graphic Objects
The Position field is worth a comment or ihe FOston Heid 1s WOU) d COMmMmMent OF two. Left-Rotated and Right-Rotated tilt text 90 degrees so. The default position, User:, sticks the text inside the frame of the graph itself. The location is measured in “virtual inches” (see “Frame & Size” on page 182) with positive numbers moving down and to the right. Frankly, sometimes it’s easiest to just stick the text any old place and then use the mouse to re-position it.
The Position field is worth a comment or ihe FOston Heid 1s WOU) d COMmMmMent OF two. Left-Rotated and Right-Rotated tilt text 90 degrees so. The default position, User:, sticks the text inside the frame of the graph itself. The location is measured in “virtual inches” (see “Frame & Size” on page 182) with positive numbers moving down and to the right. Frankly, sometimes it’s easiest to just stick the text any old place and then use the mouse to re-position it.
Shady Areas and No-Worry Lines Shady Areas and No-Worry Lines—171
Shady Areas and No-Worry Lines Shady Areas and No-Worry Lines—171
172—Chapter 6. Intimacy With Graphic Objects
172—Chapter 6. Intimacy With Graphic Objects
Templates for Success—173
Templates for Success—173
You don’t really want to re-enter the same 32 sets of shade dates every time you draw a graph, do you? Templates to the rescue! opens to the Templates & Objects section. I once made a graph with the NBER recessions marked as shades, and saved that graph in the workfile with the name RECESSIONS. Choos- ing RECESSIONS, in the Graphs as templates: scroll area, makes available all the graphic options—including shade bands—for copying into the current graph.
You don’t really want to re-enter the same 32 sets of shade dates every time you draw a graph, do you? Templates to the rescue! opens to the Templates & Objects section. I once made a graph with the NBER recessions marked as shades, and saved that graph in the workfile with the name RECESSIONS. Choos- ing RECESSIONS, in the Graphs as templates: scroll area, makes available all the graphic options—including shade bands—for copying into the current graph.
If you’d like to point out a certain observation in your data, you might want to select Draw Arrow from the right-click menu or the Proc button. The mouse cursor will turn into a crosshair. Click at the starting point, and while continuing to hold down the mouse button, drag and release at the arrow endpoint.
If you’d like to point out a certain observation in your data, you might want to select Draw Arrow from the right-click menu or the Proc button. The mouse cursor will turn into a crosshair. Click at the starting point, and while continuing to hold down the mouse button, drag and release at the arrow endpoint.
Your Data Another Sorta Way 178—Chapter 6. Intimacy With Graphic Objects You can sort a spreadsheet view. (See Sorting Things Out in Chapter 7, “Look At Your Data.”) You can also sort the data in a frozen graph. Choose Sort... from the button o1 the right-click menu.
Your Data Another Sorta Way 178—Chapter 6. Intimacy With Graphic Objects You can sort a spreadsheet view. (See Sorting Things Out in Chapter 7, “Look At Your Data.”) You can also sort the data in a frozen graph. Choose Sort... from the button o1 the right-click menu.
Hint: The Basic type page for a frozen graph that has updating off offers a limited set of options, typically far fewer than are available for a graphical view of a series or group. Graph Type
Hint: The Basic type page for a frozen graph that has updating off offers a limited set of options, typically far fewer than are available for a graphical view of a series or group. Graph Type
182—Chapter 6. Intimacy With Graphic Objects
182—Chapter 6. Intimacy With Graphic Objects
In contrast, choosing 4 inches high and 3 inches wide gives a high and nar- row frame. Options, Options, Options—183
In contrast, choosing 4 inches high and 3 inches wide gives a high and nar- row frame. Options, Options, Options—183
184—Chapter 6. Intimacy With Graphic Objects The Edit axis menu controls whether the fields below apply to the left, right, bottom or top axis. Switching the axis in the menu changes the fields in the dialog. Left and right axes For axes scaled numerically, the scaling method dropdown lets you pick between the standard linear scale, a linear scale that’s guaranteed to include zero, a log scale, and normalized data. This last scale marks the mean of the data as zero and makes one vertical unit equivalent to one standard deviation.
184—Chapter 6. Intimacy With Graphic Objects The Edit axis menu controls whether the fields below apply to the left, right, bottom or top axis. Switching the axis in the menu changes the fields in the dialog. Left and right axes For axes scaled numerically, the scaling method dropdown lets you pick between the standard linear scale, a linear scale that’s guaranteed to include zero, a log scale, and normalized data. This last scale marks the mean of the data as zero and makes one vertical unit equivalent to one standard deviation.
For the two vertical axes, the Axis scale endpoints dropdown has choices for automatic, data minimum and maximum, and user specified. Most of the time the automatic choice is fine, but once in a while you may prefer to change the scale.
For the two vertical axes, the Axis scale endpoints dropdown has choices for automatic, data minimum and maximum, and user specified. Most of the time the automatic choice is fine, but once in a while you may prefer to change the scale.
In addition, on the Data axis labels page, the Data units & label format section allows you to label your axis using scaled units or if you wish to customize the formatting of your labels.
In addition, on the Data axis labels page, the Data units & label format section allows you to label your axis using scaled units or if you wish to customize the formatting of your labels.
The Date format dropdown provides a variety of fairly self- explanatory choices, including Custom for when you want to roll your own. (See the User’s Guide.) 186—Chapter 6. Intimacy With Graphic Objects
The Date format dropdown provides a variety of fairly self- explanatory choices, including Custom for when you want to roll your own. (See the User’s Guide.) 186—Chapter 6. Intimacy With Graphic Objects
On the Attributes page, the Legend Columns entry on the left side determines how many columns are used in the legend. The default “Auto” (automatic) lets EViews use its judg- ment. Alternatively, select 1, 2, 3, etc. to specify the number of columns. Rule of thumb: the legend should be shorter than the frame. the series name itself is used. Either way, this is the spot for you to edit the legend text.
On the Attributes page, the Legend Columns entry on the left side determines how many columns are used in the legend. The default “Auto” (automatic) lets EViews use its judg- ment. Alternatively, select 1, 2, 3, etc. to specify the number of columns. Rule of thumb: the legend should be shorter than the frame. the series name itself is used. Either way, this is the spot for you to edit the legend text.
Options, Options, Options—187 Legend Aesthetics Setting the text for the legend sometimes presents a trade-off between aesthetics and informa- tion. The longer the text, the more information you can cram in. But shorter legends generally look better. Here’s a graph with a moderately long legend.
Options, Options, Options—187 Legend Aesthetics Setting the text for the legend sometimes presents a trade-off between aesthetics and informa- tion. The longer the text, the more information you can cram in. But shorter legends generally look better. Here’s a graph with a moderately long legend.
Graph Elements The Graph Elements section contains options for specific graph types. 188—Chapter 6. Intimacy With Graphic Objects
Graph Elements The Graph Elements section contains options for specific graph types. 188—Chapter 6. Intimacy With Graphic Objects
The right-most field displays the representation for each series, showing how the series will be rendered in color and how it will be rendered in black and white. The default pattern uses the colors shown in the Color column of the Attributes field for color rendering and
The right-most field displays the representation for each series, showing how the series will be rendered in color and how it will be rendered in black and white. The default pattern uses the colors shown in the Color column of the Attributes field for color rendering and
In Distinguishing Factors in the previous chapter we used Within graph category identifica- tion to have EViews automatically select visually distinct colors for different graph elements. EViews’ automatic selection produced the graph below. Fill Areas
In Distinguishing Factors in the previous chapter we used Within graph category identifica- tion to have EViews automatically select visually distinct colors for different graph elements. EViews’ automatic selection produced the graph below. Fill Areas
Using the Fill Areas page to set hatching for the first “series” (the dialog says “series,” even though the bars are really categories of a single series) produces the more visually distinc- tive version shown here. 190—Chapter 6. Intimacy With Graphic Objects
Using the Fill Areas page to set hatching for the first “series” (the dialog says “series,” even though the bars are really categories of a single series) produces the more visually distinc- tive version shown here. 190—Chapter 6. Intimacy With Graphic Objects
192—Chapter 6. Intimacy With Graphic Objects
192—Chapter 6. Intimacy With Graphic Objects
The Impact of Globalization on Intimate Graphic Activity—193 ee “wt Globalization on Intimate Graphic Activity If you do lots of similar graphs, you aren’t going to want to set the same options over and over and over and over. Templates help some. You can also set global default options through the menu Options/Graphic Defaults..., which brings up a Graph Options dialog with essentially the same sections we’ve seen already. Changes made here become the ini- tial settings for all future graphs.
The Impact of Globalization on Intimate Graphic Activity—193 ee “wt Globalization on Intimate Graphic Activity If you do lots of similar graphs, you aren’t going to want to set the same options over and over and over and over. Templates help some. You can also set global default options through the menu Options/Graphic Defaults..., which brings up a Graph Options dialog with essentially the same sections we’ve seen already. Changes made here become the ini- tial settings for all future graphs.
Quick Review? 194—Chapter 6. Intimacy With Graphic Objects Basically, EViews has a gazillion options for getting up-close and personal with graphics. Fortunately, you rarely need this level of detail because EViews has sound artistic sensibili- ties. Nonetheless, when there’s a customization you do need, it’s probably available.
Quick Review? 194—Chapter 6. Intimacy With Graphic Objects Basically, EViews has a gazillion options for getting up-close and personal with graphics. Fortunately, you rarely need this level of detail because EViews has sound artistic sensibili- ties. Nonetheless, when there’s a customization you do need, it’s probably available.
True story. I was involved in a project to analyze admissions data from the University of Washington law school. (An extract of the data, “UWLaw98.wf1”, can be found on the EViews website.) Some of my early results were really, really strange. After hours of frustra- tion I did the sensible thing and went and asked my wife’s advice. She told me:
True story. I was involved in a project to analyze admissions data from the University of Washington law school. (An extract of the data, “UWLaw98.wf1”, can be found on the EViews website.) Some of my early results were really, really strange. After hours of frustra- tion I did the sensible thing and went and asked my wife’s advice. She told me:
Sorting Things Out
Sorting Things Out
Hint: Sorting changes the order in which the data is visually displayed. The actual order in the workfile remains unchanged, so analysis is not affected. To restore the appearance to its original order, sort using Observation Order and Ascending. Describing Series—Just The Facts Please—197
Hint: Sorting changes the order in which the data is visually displayed. The actual order in the workfile remains unchanged, so analysis is not affected. To restore the appearance to its original order, sort using Observation Order and Ascending. Describing Series—Just The Facts Please—197
Was it a good idea to eliminate these two observations? This question can’t be answered by statistical analysis—you need to apply subject area knowledge. In this case, we might have chosen instead to “correct” the data by changing 39 to 3.9 and .26 to 2.6. (Although, one is left with the nagging question of whether there might really have been an applicant with a 0.26 GPA.) When we eliminated two grade observations by changing the sample, we also cut out data for other series for these two individuals. Their state of residence or LSAT scores might still be of interest, for example. There’s no right or wrong about this “side effect.” You just want to be aware that it’s happening.
Was it a good idea to eliminate these two observations? This question can’t be answered by statistical analysis—you need to apply subject area knowledge. In this case, we might have chosen instead to “correct” the data by changing 39 to 3.9 and .26 to 2.6. (Although, one is left with the nagging question of whether there might really have been an applicant with a 0.26 GPA.) When we eliminated two grade observations by changing the sample, we also cut out data for other series for these two individuals. Their state of residence or LSAT scores might still be of interest, for example. There’s no right or wrong about this “side effect.” You just want to be aware that it’s happening.
The statistic at the bottom of the panel, the Jarque-Bera, tests the hypothesis that the sample is drawn from a normal distribution. The statistic marked “Probability” is the p-value associ- ated with the Jarque-Bera. In this example, with a p-value of 0.000, the report is that it is extremely unlikely that the data follows a normal distribution.
The statistic at the bottom of the panel, the Jarque-Bera, tests the hypothesis that the sample is drawn from a normal distribution. The statistic marked “Probability” is the p-value associ- ated with the Jarque-Bera. In this example, with a p-value of 0.000, the report is that it is extremely unlikely that the data follows a normal distribution.
One-Way
One-Way
Tabulation of GPA provides lots of information. It also illustrates a com- mon problem—too many categories.
Tabulation of GPA provides lots of information. It also illustrates a com- mon problem—too many categories.
Stats By Classification Here’s a simple example. In our workfile, the vari- able WASH equals one for Washington State resi- dents and zero for everyone else. Using WASH as the classifying variable gives the results shown to the right. About 60 percent of applications (1028 out of 1639) were from out of the state, and the out of state applicants averaged a slightly higher GPA. Descriptive Statistics for GPA Categorized by values of WASH Date: 03/27/05 Time: 15:16 Sample: 8651 10367 IF GPA>1 AND GPA<5 Included observations: 1639
Stats By Classification Here’s a simple example. In our workfile, the vari- able WASH equals one for Washington State resi- dents and zero for everyone else. Using WASH as the classifying variable gives the results shown to the right. About 60 percent of applications (1028 out of 1639) were from out of the state, and the out of state applicants averaged a slightly higher GPA. Descriptive Statistics for GPA Categorized by values of WASH Date: 03/27/05 Time: 15:16 Sample: 8651 10367 IF GPA>1 AND GPA<5 Included observations: 1639
Looking at statistics by classification makes sense when the classifying variable has a small set of distinct values. When the classifying variable takes on a large number of values, it’s sometimes better to clump together values into a small number of groups or “bins.” The Group into bins if field in the lower center of the dialog lets you instruct EViews to group different values of the classifying variable into a single bin. (See Binning Control, later in this chapter.)
Looking at statistics by classification makes sense when the classifying variable has a small set of distinct values. When the classifying variable takes on a large number of values, it’s sometimes better to clump together values into a small number of groups or “bins.” The Group into bins if field in the lower center of the dialog lets you instruct EViews to group different values of the classifying variable into a single bin. (See Binning Control, later in this chapter.)
Cautionary hint for graphing multiple series: Individual series in a Group window may have NAs for different observations. As a result distribution graphs may be drawn for different samples. For example, the graph to the right makes it appear that 3-month Treasury rates are much more likely than are 1-year rates to be nearly zero. In fact what’s going on is that our sample includes 3-month, but not 1-year, rates from the Great Depression. Looking at a line graph, with the same histograms on the axis bor- der, it becomes evident that the 1-year rates enter our sample at a later date. Describing Series—Picturing the Distribution—205
Cautionary hint for graphing multiple series: Individual series in a Group window may have NAs for different observations. As a result distribution graphs may be drawn for different samples. For example, the graph to the right makes it appear that 3-month Treasury rates are much more likely than are 1-year rates to be nearly zero. In fact what’s going on is that our sample includes 3-month, but not 1-year, rates from the Great Depression. Looking at a line graph, with the same histograms on the axis bor- der, it becomes evident that the 1-year rates enter our sample at a later date. Describing Series—Picturing the Distribution—205
Views will fit any of a number of theoretical probability distributions to a series, and then lot the probability density. (You can supply the parameters of the distribution if you pre- er.) AS an example, we’ve superimposed a normal distribution on top of the GPA histo- ram.
Views will fit any of a number of theoretical probability distributions to a series, and then lot the probability density. (You can supply the parameters of the distribution if you pre- er.) AS an example, we’ve superimposed a normal distribution on top of the GPA histo- ram.
A kernel density graph is, in essence, a smoothed histogram. Often, a histogram looks choppy because the number of observations in a given bin is subject to random variation. This is particularly true when there are relatively few observations. The kernel density graph smooths the variation between nearby bins. The User’s Guide describes the various options for controlling the smoothing.
A kernel density graph is, in essence, a smoothed histogram. Often, a histogram looks choppy because the number of observations in a given bin is subject to random variation. This is particularly true when there are relatively few observations. The kernel density graph smooths the variation between nearby bins. The User’s Guide describes the various options for controlling the smoothing.
iia dienaianieaa cian einai ida aaa” “2 li aici tile-Quantile graphs and to a set of Empirical Distribution tests. Quantile-Quantile plots graph the empirical distribution of a series against a variety of theoretical probability distri- butions (e.g., normal, uniform). Empirical distribution tests provide corresponding formal tests of whether a series is drawn from a particular theoretical probability distribution. For more on these topics, see the User’s Guide—or heck, just click on the relevant menu and see what you get! Just as a histogram or kernel density plot gives an estimate of the probability density func- tion (PDF), a Cumulative Distribution plot presents an estimate of the cumulative distribu- tion function (CDF).
iia dienaianieaa cian einai ida aaa” “2 li aici tile-Quantile graphs and to a set of Empirical Distribution tests. Quantile-Quantile plots graph the empirical distribution of a series against a variety of theoretical probability distri- butions (e.g., normal, uniform). Empirical distribution tests provide corresponding formal tests of whether a series is drawn from a particular theoretical probability distribution. For more on these topics, see the User’s Guide—or heck, just click on the relevant menu and see what you get! Just as a histogram or kernel density plot gives an estimate of the probability density func- tion (PDF), a Cumulative Distribution plot presents an estimate of the cumulative distribu- tion function (CDF).
Opening the boxplot Graph Options dialog. A boxplot of GPA using EViews’ defaults is shown here.
Opening the boxplot Graph Options dialog. A boxplot of GPA using EViews’ defaults is shown here.
Statistics don’t lie, but they can mislead. In this plot, the upper staple for non-residents is higher than the staple for residents. A little investigation shows that, while GPA is measured on a four point scale, one observation for a non-residents was recorded 4.1. Because some schools use something other than a four point scale, we don’t know if this GPA is an error or not. But we probably shouldn’t conclude anything about the difference between in- and out-of-state applicants based on this one data point.
Statistics don’t lie, but they can mislead. In this plot, the upper staple for non-residents is higher than the staple for residents. A little investigation shows that, while GPA is measured on a four point scale, one observation for a non-residents was recorded 4.1. Because some schools use something other than a four point scale, we don’t know if this GPA is an error or not. But we probably shouldn’t conclude anything about the difference between in- and out-of-state applicants based on this one data point.
Tests By Classification Buzzword hint: We’d say that the average UW applicant LSAT is statistically signifi- cantly different from the average of all test takers.
Tests By Classification Buzzword hint: We’d say that the average UW applicant LSAT is statistically signifi- cantly different from the average of all test takers.
Test for Equality of Means of GPA Categorized by values of WASH Date: 12/06/06 Time: 14:08 Sample: 8651 10367 IF (GPA>1 AND GPA<5 AND LSAT=100) Included observations: 1638 While the difference between in-state and out-of-state GPAs is very small, statistically it’s highly significant. *Test allows for unequal cell variances
Test for Equality of Means of GPA Categorized by values of WASH Date: 12/06/06 Time: 14:08 Sample: 8651 10367 IF (GPA>1 AND GPA<5 AND LSAT=100) Included observations: 1638 While the difference between in-state and out-of-state GPAs is very small, statistically it’s highly significant. *Test allows for unequal cell variances
People use correlations a lot more because correla- tions are unit-free, while the units of covariances depend on the units of the underlying series.
People use correlations a lot more because correla- tions are unit-free, while the units of covariances depend on the units of the underlying series.
Describing Groups—Just the Facts—Putting It Together—215 Here, we see the correlation for the two series in the group. Surprisingly, LSAT and GPA are not all that highly correlated. A correlation of only 0.307 means that the information in LSAT is not redun- dant with the information in GPA, which is why law schools look at both test scores and grades.
Describing Groups—Just the Facts—Putting It Together—215 Here, we see the correlation for the two series in the group. Surprisingly, LSAT and GPA are not all that highly correlated. A correlation of only 0.307 means that the information in LSAT is not redun- dant with the information in GPA, which is why law schools look at both test scores and grades.
Hint: The intersection of a row and column is called a cell. For example, there are 35\ applicants in the High grade/Low test score cell.
Hint: The intersection of a row and column is called a cell. For example, there are 35\ applicants in the High grade/Low test score cell.
Table Interpretation The bottom row reports the totals for each column and the right-most column reports the totals for each row. The Total-Total, bottom right, is the number of observations used in the table.
Table Interpretation The bottom row reports the totals for each column and the right-most column reports the totals for each row. The Total-Total, bottom right, is the number of observations used in the table.
For this example, we’re going to model currency growth as a linear function of a time trend, lagged currency growth, and a different constant for each month of the year. We need to estimate an equation for this model before we can make a forecast. Here’s the relevant command: Just Push the Forecast Button
For this example, we’re going to model currency growth as a linear function of a time trend, lagged currency growth, and a different constant for each month of the year. We need to estimate an equation for this model before we can make a forecast. Here’s the relevant command: Just Push the Forecast Button
| Equation: UNTITLED Workfile: CURRENCY::Currcirt The estimation results look fine. You’re done. The forecasts for G are stored in the series GF.
| Equation: UNTITLED Workfile: CURRENCY::Currcirt The estimation results look fine. You’re done. The forecasts for G are stored in the series GF.
Let’s review a little forecasting theory and then see where EViews fits in
Let’s review a little forecasting theory and then see where EViews fits in
When you click the button EViews does the same calculation—just faster and with some extra doo-dads available in the output.
When you click the button EViews does the same calculation—just faster and with some extra doo-dads available in the output.
Setting the Sample in the Forecast Dialog
Setting the Sample in the Forecast Dialog
Chow Breakpoint Test 1950M01 Null Hypothesis: No breaks at specified breakpoints Varying regressors: All equation variables Equation Sample: 1917M10 2000M12 If we define an “alternate history” with:
Chow Breakpoint Test 1950M01 Null Hypothesis: No breaks at specified breakpoints Varying regressors: All equation variables Equation Sample: 1917M10 2000M12 If we define an “alternate history” with:
The Forecast evaluation option generates a small table with a variety of statistics for comparing fore- cast and actual values. The Root Mean Squared Error (or RMSE) is the standard deviation of the forecast errors. (See the User’s Guide for explanations of the other statistics.) Forecast: GHEIN_DYN Actual: G Forecast sample: 2001M01 2005M04 Included observations: 52
The Forecast evaluation option generates a small table with a variety of statistics for comparing fore- cast and actual values. The Root Mean Squared Error (or RMSE) is the standard deviation of the forecast errors. (See the User’s Guide for explanations of the other statistics.) Forecast: GHEIN_DYN Actual: G Forecast sample: 2001M01 2005M04 Included observations: 52
Glance back at the data plot which opened the chapter. It shows an enormous increase in currency holdings right before the turn of the millennium and a huge drop imme- diately thereafter. This means that a dynamic forecast starting in the beginning of 2000 uses an anoma- lous value for lagged G, a problem which is carried forward. In con- trast, a static forecast should only have difficulty at the beginning of the forecast period, since thereafter actual lagged G picks up non-anom- alous data.
Glance back at the data plot which opened the chapter. It shows an enormous increase in currency holdings right before the turn of the millennium and a huge drop imme- diately thereafter. This means that a dynamic forecast starting in the beginning of 2000 uses an anoma- lous value for lagged G, a problem which is carried forward. In con- trast, a static forecast should only have difficulty at the beginning of the forecast period, since thereafter actual lagged G picks up non-anom- alous data.
Forecasting Beneath the Surface—235 Forecasting Beneath the Surface Sometimes the variable you want to forecast isn’t quite the variable on the left of your esti- mating equation. There are often statistical or economic modeling reasons for estimating a transformed version of the variable you really care about. Two common examples are using logs (rather than levels) of the variable of interest, and using first differences of the variable of interest.
Forecasting Beneath the Surface—235 Forecasting Beneath the Surface Sometimes the variable you want to forecast isn’t quite the variable on the left of your esti- mating equation. There are often statistical or economic modeling reasons for estimating a transformed version of the variable you really care about. Two common examples are using logs (rather than levels) of the variable of interest, and using first differences of the variable of interest.
Hint: If you use an expression, for example “1200*dlog(curr)”, rather than a named auto-series as the dependent variable, you get pretty much the same choices, although the dialogs look a little different. Forecasting Beneath the Surface—237
Hint: If you use an expression, for example “1200*dlog(curr)”, rather than a named auto-series as the dependent variable, you get pretty much the same choices, although the dialogs look a little different. Forecasting Beneath the Surface—237
Quick Review—Forecasting
Quick Review—Forecasting
Hint: The Workfile Create dialog always gives you the opportunity to name the page being created, even if you’re setting up a separate workfile rather than a page as in this example. If you leave the Page field blank, EViews assigns “Untitled” or “Untitled1,” etc., for the page name. Creating New Pages—241
Hint: The Workfile Create dialog always gives you the opportunity to name the page being created, even if you’re setting up a separate workfile rather than a page as in this example. If you leave the Page field blank, EViews assigns “Untitled” or “Untitled1,” etc., for the page name. Creating New Pages—241
Copy/Extract from Current Page copies data from the active workfile page into a new page (or a standalone workfile, if you prefer). The command is straightforward once you get over the name. This menu is accessed by clicking on the New Page tab, but “Current Page” doesn’t mean the new page attached to the tab you just clicked on—it means the currently active page. Just pretend that the command is named Copy/Extract from Active Page.
Copy/Extract from Current Page copies data from the active workfile page into a new page (or a standalone workfile, if you prefer). The command is straightforward once you get over the name. This menu is accessed by clicking on the New Page tab, but “Current Page” doesn’t mean the new page attached to the tab you just clicked on—it means the currently active page. Just pretend that the command is named Copy/Extract from Active Page.
Copy-and-Paste and Drag-and-Drop
Copy-and-Paste and Drag-and-Drop
Up-Frequency Conversions
Up-Frequency Conversions
Multiple Frequencies—Multiple Pages—247
Multiple Frequencies—Multiple Pages—247
As background, let’s take a look at quarterly GDP values for the year 2004 and the monthly values manufactured by EViews. We can see that EViews copied the value of GDP in a given quarter into each month of that quarter. That’s a reasonable thing to do. If all we know was that GDP in the first quarter was running at a rate of a trillion dollars a year, then saying that GDP was running at a trillion a year in January, a trillion a year in February, and a trillion a year in March seems like a good Start.
As background, let’s take a look at quarterly GDP values for the year 2004 and the monthly values manufactured by EViews. We can see that EViews copied the value of GDP in a given quarter into each month of that quarter. That’s a reasonable thing to do. If all we know was that GDP in the first quarter was running at a rate of a trillion dollars a year, then saying that GDP was running at a trillion a year in January, a trillion a year in February, and a trillion a year in March seems like a good Start.
Default Frequency Conversions Hint: First observation and Last observation are especially useful in the analysis of financial price data, where point-in-time data are often preferred over time aggregated data. EViews has applied the default high-to-low frequency conver- sion procedure, which is to average the monthly observations within each quarter. The meanings of the high-to-low frequency conversion options are pretty straightforward. Sum observa- tions adds up the monthly observations within a quarter. If we had sales for January, February, and March we would use a Sum observations conversion to get total first quarter sales. First, Last, Max, and Min Observation all pick out one month within the quarter ar the value for the quarterly value. Specified in series Average observations Sum observations First observation Last observations Max observation Min observation No down conversions the value for the quarterly value.
Default Frequency Conversions Hint: First observation and Last observation are especially useful in the analysis of financial price data, where point-in-time data are often preferred over time aggregated data. EViews has applied the default high-to-low frequency conver- sion procedure, which is to average the monthly observations within each quarter. The meanings of the high-to-low frequency conversion options are pretty straightforward. Sum observa- tions adds up the monthly observations within a quarter. If we had sales for January, February, and March we would use a Sum observations conversion to get total first quarter sales. First, Last, Max, and Min Observation all pick out one month within the quarter ar the value for the quarterly value. Specified in series Average observations Sum observations First observation Last observations Max observation Min observation No down conversions the value for the quarterly value.
Paste Special Nothing limits copy-and-paste or copy-and-paste-special to a single series. The Pattern field lets you specify a general pattern for changing the name of the pasted series. For example, the pattern “*_quarterly” would paste the series X, Y, and Z as X_quarterly, Y_quarterly, and Z_quarterly. For further discussion, see the User’s Guide.
Paste Special Nothing limits copy-and-paste or copy-and-paste-special to a single series. The Pattern field lets you specify a general pattern for changing the name of the pasted series. For example, the pattern “*_quarterly” would paste the series X, Y, and Z as X_quarterly, Y_quarterly, and Z_quarterly. For further discussion, see the User’s Guide.
Or, you might not need to create a link object in your workfile page at all. Anywhere that you can use a series, you can also use a series from another workfile page. If you are creat-
Or, you might not need to create a link object in your workfile page at all. Anywhere that you can use a series, you can also use a series from another workfile page. If you are creat-
Break Links - convert into ordinary series button, detaches the specified links from their source data and converts them into regular numeric or alpha series using the current value: in all links or just those you list. Unlinking If frequency conversion between the pages is necessary, default frequency conversion will be used.
Break Links - convert into ordinary series button, detaches the specified links from their source data and converts them into regular numeric or alpha series using the current value: in all links or just those you list. Unlinking If frequency conversion between the pages is necessary, default frequency conversion will be used.
There are days when computers are incredibly annoying. The connection between pages is obvious to us, but not to the computer, because the observations aren’t quite parallel in the two pages. The infant mortality data includes an observation for the District of Columbia; The workfile “Infant Mortality Rate.wfl” holds two pages with data by state. The page Mor- tality contains infant mortality rates and the page Revenue contains per capita revenue. Is there a connection between the two? Excerpts of the data look like:
There are days when computers are incredibly annoying. The connection between pages is obvious to us, but not to the computer, because the observations aren’t quite parallel in the two pages. The infant mortality data includes an observation for the District of Columbia; The workfile “Infant Mortality Rate.wfl” holds two pages with data by state. The page Mor- tality contains infant mortality rates and the page Revenue contains per capita revenue. Is there a connection between the two? Excerpts of the data look like:
Click on the Mortality tab to activate the Mortal- ity page. Then create a new link named REV using the menu Object/New Object/Series Link. The object Jrev appears in the workfile win- dow with a pink background to indicate a link and a question mark showing that the link hasn’t been specified. Double-clicking BJ rev opens a view with an error message indicating that the properties of the link haven’t been specified yet.
Click on the Mortality tab to activate the Mortal- ity page. Then create a new link named REV using the menu Object/New Object/Series Link. The object Jrev appears in the workfile win- dow with a pink background to indicate a link and a question mark showing that the link hasn’t been specified. Double-clicking BJ rev opens a view with an error message indicating that the properties of the link haven’t been specified yet.
Hint: You may find it more intuitive to think of the Link to field as the “Link from” field. Remember that you’re specifying the data source here. The destination is always the active page.
Hint: You may find it more intuitive to think of the Link to field as the “Link from” field. Remember that you’re specifying the data source here. The destination is always the active page.
Specifying A Page By Identifier Series Matching When The Identifiers Are Really Different—259
Specifying A Page By Identifier Series Matching When The Identifiers Are Really Different—259
Our next step is to create a page holding data aggregated to the state level. We want our new page to contain one observa- tion for each of the states observed in the individual data. Clicking on the New Page tab brings up a menu including the choice Specify by Identifier Series..., which (not surpris- ingly) is just what we need to specify an identifier series for a new page. In our original page the state identifier is in the series GMSTCEN, so that’s what we’ll use as the identifier for the Specify by Frequency/Range ... Specify by Identifier Series ... Copy/Extract from Current Page > Load Workfile Page... Paste from Clipboard as Page
Our next step is to create a page holding data aggregated to the state level. We want our new page to contain one observa- tion for each of the states observed in the individual data. Clicking on the New Page tab brings up a menu including the choice Specify by Identifier Series..., which (not surpris- ingly) is just what we need to specify an identifier series for a new page. In our original page the state identifier is in the series GMSTCEN, so that’s what we’ll use as the identifier for the Specify by Frequency/Range ... Specify by Identifier Series ... Copy/Extract from Current Page > Load Workfile Page... Paste from Clipboard as Page
The new page opens containing just the series GMSTCEN. Okay, the new page also contains FM11X—but that’s only because FM11X is a value map holding the names of each state.
The new page opens containing just the series GMSTCEN. Okay, the new page also contains FM11X—but that’s only because FM11X is a value map holding the names of each state.
method field, switch to Number of obs to get a count of how many observations are being used for each series.
method field, switch to Number of obs to get a count of how many observations are being used for each series.
ere’s what happened in the mean merge. elie he contraction computed the mean for each ries separately. The Current Population Sur- »y isn’t limited to workers, so the state-by- ate means have been computed as a fraction ‘the population. We probably wanted only ose who are working. What’s more, the vari- le LNWAGE is coded as NA for anyone who esn’t report a positive salary, including all on-workers. As a result, the state-by-state ieans for LNWAGE were computed using roughly 25 | her variables. Hint: We could have edited the link specifications, but since we had several series, it was faster to just toss the links and start over.
ere’s what happened in the mean merge. elie he contraction computed the mean for each ries separately. The Current Population Sur- »y isn’t limited to workers, so the state-by- ate means have been computed as a fraction ‘the population. We probably wanted only ose who are working. What’s more, the vari- le LNWAGE is coded as NA for anyone who esn’t report a positive salary, including all on-workers. As a result, the state-by-state ieans for LNWAGE were computed using roughly 25 | her variables. Hint: We could have edited the link specifications, but since we had several series, it was faster to just toss the links and start over.
| Equation: UNTITLED Workfile: CPSMAR2004EXTRACT::ByStatet AL UIE stdl€-Dy-stdat€ L€steoslOl, ee the interpretation of the union coefficient has changed. Because UNION is measured as a fraction, the regression now tells us that for each one percentage point increase in the unionization rate, the average wage rises two and a half percent.
| Equation: UNTITLED Workfile: CPSMAR2004EXTRACT::ByStatet AL UIE stdl€-Dy-stdat€ L€steoslOl, ee the interpretation of the union coefficient has changed. Because UNION is measured as a fraction, the regression now tells us that for each one percentage point increase in the unionization rate, the average wage rises two and a half percent.
The first few observations in the Cps page are shown to the right. Notice that the fourth and fifth person are both from Cor necticut. Even though the fourth person is a union member and the fifth person isn’t, the have the same value of AV_UNION. We’ve succeeded in attaching the state-wide averag' unionization rate to each individual observa- tion.
The first few observations in the Cps page are shown to the right. Notice that the fourth and fifth person are both from Cor necticut. Even though the fourth person is a union member and the fifth person isn’t, the have the same value of AV_UNION. We’ve succeeded in attaching the state-wide averag' unionization rate to each individual observa- tion.
For the pooled data on the left, we see the first few observations for population for Canada and the U.S., each in its own series. The two series have (intentionally and usefully) similar names, but nothing is fundamentally different from what we’ve seen before. Here’s a quick peek at the data. The picture on the right shows the panel approach, which introduces a kind of structure in the workfile that we haven’t seen before. The Range field now reads “1950 2000 x 2.” The data are still annual from 1950 through 2000, but the workfile is structured to contain two cross sections (Canada and the U.S.). All the population measurements are in the single series POP.
For the pooled data on the left, we see the first few observations for population for Canada and the U.S., each in its own series. The two series have (intentionally and usefully) similar names, but nothing is fundamentally different from what we’ve seen before. Here’s a quick peek at the data. The picture on the right shows the panel approach, which introduces a kind of structure in the workfile that we haven’t seen before. The Range field now reads “1950 2000 x 2.” The data are still annual from 1950 through 2000, but the workfile is structured to contain two cross sections (Canada and the U.S.). All the population measurements are in the single series POP.
Quick (P)review If you have a cross-section of time series, put them into a pool. If you have repeated obser- vations on cross-section elements, set up a panel. We'll spend the next two chapters answering this question. The big answer is that you can control for common elements across observations—or not—as you choose. The smaller answer is that all sorts of data manipulation are made easier because EViews understands how different observations are tied together.
Quick (P)review If you have a cross-section of time series, put them into a pool. If you have repeated obser- vations on cross-section elements, set up a panel. We'll spend the next two chapters answering this question. The big answer is that you can control for common elements across observations—or not—as you choose. The smaller answer is that all sorts of data manipulation are made easier because EViews understands how different observations are tied together.
The figure to the right shows observations 1597 through 1610, which happen to be the last few observations for the Central African Republic and the first few observations for Canada. To us humans, it’s clear that these are observations for i = CAF,CAN and for t = 1994...2000 and then starting over with ¢ = 1950... . In order to set up a panel structure, we need to share this kind of understanding with EViews.
The figure to the right shows observations 1597 through 1610, which happen to be the last few observations for the Central African Republic and the first few observations for Canada. To us humans, it’s clear that these are observations for i = CAF,CAN and for t = 1994...2000 and then starting over with ¢ = 1950... . In order to set up a panel structure, we need to share this kind of understanding with EViews.
Range field in the top panel of the Setting Up Panel Data—277
Range field in the top panel of the Setting Up Panel Data—277
I Workfile: PWI61EXTRACT - (czleviewstdatalpwt6textract.... E)G)®) Let’s take another look at our data, this time as displayed in the panel workfile. Now the obs column correctly identifies each data point with both country code and year. That’s about all you need to know to set up a panel workfile. One more option is worth mentioning. Should you “balance?”
I Workfile: PWI61EXTRACT - (czleviewstdatalpwt6textract.... E)G)®) Let’s take another look at our data, this time as displayed in the panel workfile. Now the obs column correctly identifies each data point with both country code and year. That’s about all you need to know to set up a panel workfile. One more option is worth mentioning. Should you “balance?”
us an unbalanced panel. EViews default (if you leave the check box Balance between starts
us an unbalanced panel. EViews default (if you leave the check box Balance between starts
But is the effect of population growth important? We can try to get a better handle on this by comparing a couple of countries; let’s use the Central African Republic and Canada, since they’re at opposite ends of the development spectrum. Set the sample using:
But is the effect of population growth important? We can try to get a better handle on this by comparing a couple of countries; let’s use the Central African Republic and Canada, since they’re at opposite ends of the development spectrum. Set the sample using:
The average value of population growth was 2.2 percent per year in the Central African Republic and 1.6 percent in Canada. If we multiply the difference in population growth rates, 0.008, by the estimated regression coefficient, -901, we predict that relative GDP in the Central African Repub- lic should be 7 percentage points lower than in Canada. Population growth appears to have a very large effect.
The average value of population growth was 2.2 percent per year in the Central African Republic and 1.6 percent in Canada. If we multiply the difference in population growth rates, 0.008, by the estimated regression coefficient, -901, we predict that relative GDP in the Central African Repub- lic should be 7 percentage points lower than in Canada. Population growth appears to have a very large effect.
Econometric digression: The regression output includes a hint that something funky is going on. The Durbin-Watson statistic (see Chapter 13, “Serial Correlation—Friend or Foe?”) indicates very, very high serial correlation. This suggests that if the error for a country in one year is positive then it’s positive in all years, and if it’s negative once then it’s always negative. High serial correlation in this context provides a hint that we've left out country specific information.
Econometric digression: The regression output includes a hint that something funky is going on. The Durbin-Watson statistic (see Chapter 13, “Serial Correlation—Friend or Foe?”) indicates very, very high serial correlation. This suggests that if the error for a country in one year is positive then it’s positive in all years, and if it’s negative once then it’s always negative. High serial correlation in this context provides a hint that we've left out country specific information.
We can take a look at the estimated values of the fixed effects for each country by looking at the Fixed/Ran- lom Effects/Cross-section Effects view. The reported
We can take a look at the estimated values of the fixed effects for each country by looking at the Fixed/Ran- lom Effects/Cross-section Effects view. The reported
| Equation: UNTITLED. Workfile: PWI61EXTRACT::Pwt61'
| Equation: UNTITLED. Workfile: PWI61EXTRACT::Pwt61'
Pretty Panel Pictures
Pretty Panel Pictures
More Panel Estimation Techniques Hint: EViews doesn’t produce a legend for this sort of graph by default. You can dou- ble-click on the graph, go to the legend portion of the dialog, and tell it do so.
More Panel Estimation Techniques Hint: EViews doesn’t produce a legend for this sort of graph by default. You can dou- ble-click on the graph, go to the legend portion of the dialog, and tell it do so.
We can look at this issue by including dummy variables for the 51 states (DC too, eh?). This can be done directly (and we’ll do it directly in the next section) but it can be very conve- nient to pretend that each state identifies a cross section in a panel so that we can use EViews panel estimation tools. In other words, let’s fake a panel.
We can look at this issue by including dummy variables for the 51 states (DC too, eh?). This can be done directly (and we’ll do it directly in the next section) but it can be very conve- nient to pretend that each state identifies a cross section in a panel so that we can use EViews panel estimation tools. In other words, let’s fake a panel.
Fixed Effects With and Without the Social Contrivance of Panel Structure Fixed Effects With and Without the Social Contrivance of Panel Structure—287 EViews provides a large set of features designed for panel data, but fixed effects estimation is the most important. In terms of econometrics, specifying fixed effects in a linear regres- sion is a fancy way of including a dummy variable for each group (country or state in the examples in this chapter). You’re free to include these dummy variables manually, if you wish.
Fixed Effects With and Without the Social Contrivance of Panel Structure Fixed Effects With and Without the Social Contrivance of Panel Structure—287 EViews provides a large set of features designed for panel data, but fixed effects estimation is the most important. In terms of econometrics, specifying fixed effects in a linear regres- sion is a fancy way of including a dummy variable for each group (country or state in the examples in this chapter). You’re free to include these dummy variables manually, if you wish.
ls Inwage c ed age asian @expand(gmstcen, @droplast)
ls Inwage c ed age asian @expand(gmstcen, @droplast)
Getting Your Feet Wet
Getting Your Feet Wet
Reverse causation alert: There’s good reason to believe that countries becoming richer leads to lower population growth. Thus there’s a real issue of whether we’re picking up the effect of output on population growth rather than population growth on output. The issue is real, but it hasn’t got anything to do with illustrating the use of pools, so we won’t worry about it further.
Reverse causation alert: There’s good reason to believe that countries becoming richer leads to lower population growth. Thus there’s a real issue of whether we’re picking up the effect of output on population growth rather than population growth on output. The issue is real, but it hasn’t got anything to do with illustrating the use of pools, so we won’t worry about it further.
Dependent Variable: ¥? Method: Pooled Least Squares Date: 07/17/09 Time: 14:40 Sample (adjusted): 1951 2000 Included observations: 50 after adjustment Cross-sections included: 6 Total pool (unbalanced) observations: 280
Dependent Variable: ¥? Method: Pooled Least Squares Date: 07/17/09 Time: 14:40 Sample (adjusted): 1951 2000 Included observations: 50 after adjustment Cross-sections included: 6 Total pool (unbalanced) observations: 280
Dependent Variable: ¥? Method: Pooled Least Squares Date: 07/17/09 Time: 14:41 Sample (adjusted): 1951 2000 Included observations: 50 after adjustments Cross-sections included: 6 Total pool (unbalanced) observations: 280
Dependent Variable: ¥? Method: Pooled Least Squares Date: 07/17/09 Time: 14:41 Sample (adjusted): 1951 2000 Included observations: 50 after adjustments Cross-sections included: 6 Total pool (unbalanced) observations: 280
Everyone Into the Pool May Not Be Fun Now we're asking for a separate intercept for each country. The estimated effect of population growth is negative as we had expected. And the increased sam- ple size has raised the t-statistic on population growth from 3 to 5.
Everyone Into the Pool May Not Be Fun Now we're asking for a separate intercept for each country. The estimated effect of population growth is negative as we had expected. And the increased sam- ple size has raised the t-statistic on population growth from 3 to 5.
Fixed Effects
Fixed Effects
Stacked View
Stacked View
Unstacked View
Unstacked View
The default Data Organization is Playing in the Pool—Data—299
The default Data Organization is Playing in the Pool—Data—299
4mpirical aside: If you’re following along on the computer, you can scroll the output o see that three of the countries in the pooled sample have population growth stan- lard deviations much lower than the U.S. and three have standard deviations a little ibove that of the U.S.
4mpirical aside: If you’re following along on the computer, you can scroll the output o see that three of the countries in the pooled sample have population growth stan- lard deviations much lower than the U.S. and three have standard deviations a little ibove that of the U.S.
Playing in the Pool—Data—301
Playing in the Pool—Data—301
Much of our analysis on the pool has used percentage population growth, measured as D(LOG(POP?)). We might want to generate this as a new series for each country. Manually, we could give six commands of the form:
Much of our analysis on the pool has used percentage population growth, measured as D(LOG(POP?)). We might want to generate this as a new series for each country. Manually, we could give six commands of the form:
To delete a pile of series, choose Delete Pool
To delete a pile of series, choose Delete Pool
Residuals are supposed to be centered on zero. The second thing you’ll notice is that the residuals for Canada are all strongly positive, those for Germany are mostly positive and Italy’s residuals are nearly all negative, so the residuals are not centered on zero. That’s a hint that the country equations should have different intercepts. Return to the first pooled estimate in the chapter, the one with a common intercept for all series. Pick the menu View/Residuals/Graphs. The first thing you'll notice is that squeez- ing six graphs into one window makes for some pretty tiny graphs.
Residuals are supposed to be centered on zero. The second thing you’ll notice is that the residuals for Canada are all strongly positive, those for Germany are mostly positive and Italy’s residuals are nearly all negative, so the residuals are not centered on zero. That’s a hint that the country equations should have different intercepts. Return to the first pooled estimate in the chapter, the one with a common intercept for all series. Pick the menu View/Residuals/Graphs. The first thing you'll notice is that squeez- ing six graphs into one window makes for some pretty tiny graphs.
Grabbing the Residual Series if six graphs in one window i looking a little hard to read, think what sixty-six graphs would look like! Proc/Make More Pool Estimation—305
Grabbing the Residual Series if six graphs in one window i looking a little hard to read, think what sixty-six graphs would look like! Proc/Make More Pool Estimation—305
Nomenclature hint: SUR stands for Seemingly Unrelated Regression.
Nomenclature hint: SUR stands for Seemingly Unrelated Regression.
Since we have ordinary series with conveniently chosen names, load in the spreadsheet in the usual way, create a pool object with suffixes CAN, FRA, etc., and bob’s your uncle. Here’s an excerpt of an Excel spreadsheet with unstacked data.
Since we have ordinary series with conveniently chosen names, load in the spreadsheet in the usual way, create a pool object with suffixes CAN, FRA, etc., and bob’s your uncle. Here’s an excerpt of an Excel spreadsheet with unstacked data.
Importing Stacked Data—The Direct Method
Importing Stacked Data—The Direct Method
Importing Stacked Data—The Indirect Method
Importing Stacked Data—The Indirect Method
The pool feature lets you analyze multiple series observed for the same variable, such as GDP series for a number of countries. You can pool the data in a regression with common Exporting Stacked Data—A Little Indirection Here Too Hint: EViews includes the “?” in the series name in the output file. You might choose to manually delete the “2” in the exported file to improve the appearance of the out- put.
The pool feature lets you analyze multiple series observed for the same variable, such as GDP series for a number of countries. You can pool the data in a regression with common Exporting Stacked Data—A Little Indirection Here Too Hint: EViews includes the “?” in the series name in the output file. You might choose to manually delete the “2” in the exported file to improve the appearance of the out- put.
Exporting Stacked Data
Exporting Stacked Data
ls logvol c @trend @trend*2 d(log(close(-1))) The next commands save the fitted values and the residuals, and then open a group—which we changed to a line graph and then prettied up a little.
ls logvol c @trend @trend*2 d(log(close(-1))) The next commands save the fitted values and the residuals, and then open a group—which we changed to a line graph and then prettied up a little.
Every regression estimate comes with views to make looking at residuals easy. If the errors are serially correlated, then a large residual should generally be followed by another large residual; a small residual is likely to be followed by another small residual; and positive followed by positive and negative by negative. Clicking the button brings up the Actual, Fitte
Every regression estimate comes with views to make looking at residuals easy. If the errors are serially correlated, then a large residual should generally be followed by another large residual; a small residual is likely to be followed by another small residual; and positive followed by positive and negative by negative. Clicking the button brings up the Actual, Fitte
Econometric hint: We’re treating serial correlation as a statistical issue. Sometimes serial correlation is a hint of misspecification. Although it’s not something we'll inves- tigate further, that’s probably the case here.
Econometric hint: We’re treating serial correlation as a statistical issue. Sometimes serial correlation is a hint of misspecification. Although it’s not something we'll inves- tigate further, that’s probably the case here.
Durbin-Watson Statistic Visual checks provide a great deal of information, but you’ll probably want to follow up with one or more formal statistical tests for serial correlation. EViews provides three test sta- tistics: the Durbin-Watson, the Breusch-Godfrey, and the Ljung-Box Q-statistic.
Durbin-Watson Statistic Visual checks provide a great deal of information, but you’ll probably want to follow up with one or more formal statistical tests for serial correlation. EViews provides three test sta- tistics: the Durbin-Watson, the Breusch-Godfrey, and the Ljung-Box Q-statistic.
Breusch-Godfrey Statistic
Breusch-Godfrey Statistic
Breusch-Godfrey Serial Correlation LM Test: Ljung-Box Q-statistic
Breusch-Godfrey Serial Correlation LM Test: Ljung-Box Q-statistic
A different approach to checking for serial correlation is to plot the correlation of the resid- ual with the residual lagged once, the residual with the residual lagged twice, and so on. As we saw above in The Correlogram, this plot is called the correlogram of the residuals. If there is no serial correlation then correlations should all be zero, except for random fluctuation. To see the correlogram, choose Residual Diagnostics/Correlogram - Q-statistics... from the View] menu. A small dialog pops open allowing you to specify the number of correlations to show.
A different approach to checking for serial correlation is to plot the correlation of the resid- ual with the residual lagged once, the residual with the residual lagged twice, and so on. As we saw above in The Correlogram, this plot is called the correlogram of the residuals. If there is no serial correlation then correlations should all be zero, except for random fluctuation. To see the correlogram, choose Residual Diagnostics/Correlogram - Q-statistics... from the View] menu. A small dialog pops open allowing you to specify the number of correlations to show.
Equations that include ARMA parameters have an ARMA Structure... view which brings up a dialog offering four diagnostics. We’ll take a look at the Correlogram view here and the Impulse Response view in the next section. Correcting for Serial Correlation—325
Equations that include ARMA parameters have an ARMA Structure... view which brings up a dialog offering four diagnostics. We’ll take a look at the Correlogram view here and the Impulse Response view in the next section. Correcting for Serial Correlation—325
Another Way to Look at the ARMA Coefficients
Another Way to Look at the ARMA Coefficients
The solid line (theoretical) and the top of the spikes (empirical) don’t match up very well, do they? The pattern suggests that an AR(1) isn’t a good enough specification, which we already suspected from other evidence. The Impulse Response Function
The solid line (theoretical) and the top of the spikes (empirical) don’t match up very well, do they? The pattern suggests that an AR(1) isn’t a good enough specification, which we already suspected from other evidence. The Impulse Response Function
Nomenclature Hint: The theoretical correlogram corresponding to the estimated ARMA parameters is sometimes called the Autocorrelation Function or ACF.
Nomenclature Hint: The theoretical correlogram corresponding to the estimated ARMA parameters is sometimes called the Autocorrelation Function or ACF.
ARMA and ARIMA Models
ARMA and ARIMA Models
Who Put the | in ARIMA? standard solution to this problem is to build a model of the first difference of the variable
Who Put the | in ARIMA? standard solution to this problem is to build a model of the first difference of the variable
Unit Root Tests Hint: The first difference of LOGVOL can be written in two ways: D(LOGVOL) or LOGVOL-LOGVOL(-1). The two are equivalent. Hint: If you know dy, = y;-— Yo» @yy = Yo — y; » etc., then you can find y, by adding dy, to yj. You can find y, by adding dy, and dy, to y,, and so forth. Adding up the first differences is the source of the term “integrated.”
Unit Root Tests Hint: The first difference of LOGVOL can be written in two ways: D(LOGVOL) or LOGVOL-LOGVOL(-1). The two are equivalent. Hint: If you know dy, = y;-— Yo» @yy = Yo — y; » etc., then you can find y, by adding dy, to yj. You can find y, by adding dy, and dy, to y,, and so forth. Adding up the first differences is the source of the term “integrated.”
So with that build up, here’s how to estimate an ARIMA(1,1,1) model in EViews: ARIMA Estimation
So with that build up, here’s how to estimate an ARIMA(1,1,1) model in EViews: ARIMA Estimation
Dependent Variable: D(LOGYOL) Method: Least Squares Date: 07/17/09 Time: 14:16 Sample (adjusted): 188803 200401 Included observations: 463 after adjustment Convergence achieved after 6 iterations ha Darbract 1000990 332—Chapter 13. Serial Correlation—Friend or Foe?
Dependent Variable: D(LOGYOL) Method: Least Squares Date: 07/17/09 Time: 14:16 Sample (adjusted): 188803 200401 Included observations: 463 after adjustment Convergence achieved after 6 iterations ha Darbract 1000990 332—Chapter 13. Serial Correlation—Friend or Foe?
Quick Review
Quick Review
Dependent Variable: G Method: Least Squares Date: 12/21/06 Time: 15:06 Sample (adjusted): 1917M10 2005mM04 Included observations: 1045 after adjustments
Dependent Variable: G Method: Least Squares Date: 12/21/06 Time: 15:06 Sample (adjusted): 1917M10 2005mM04 Included observations: 1045 after adjustments
336—Chapter 14. A Taste of Advanced Estimation
336—Chapter 14. A Taste of Advanced Estimation
The Weighted Option
The Weighted Option
Dependent Variable: G Method: Least Squares Date: 07/23/09 Time: 12:22 Sample (adjusted): 1917M10 2005mM04 Included observations: 1045 after adjustments White heteroskedasticity-consistent standard errors & covariar
Dependent Variable: G Method: Least Squares Date: 07/23/09 Time: 12:22 Sample (adjusted): 1917M10 2005mM04 Included observations: 1045 after adjustments White heteroskedasticity-consistent standard errors & covariar
sroup. As an example, we’ll trumpet the White results. Heteroskedasticity Robust Standard Errors
sroup. As an example, we’ll trumpet the White results. Heteroskedasticity Robust Standard Errors
Nonlinear Least Squares Nonlinear Least Squares—341 Hint: It may look curious that C(1), C(2), etc. seem to be labeled R1, R2, and so on. R1 and R2 are generic labels for row 1 and row 2 in a coefficient vector. The same setup is used generally for displaying vectors and matrices.
Nonlinear Least Squares Nonlinear Least Squares—341 Hint: It may look curious that C(1), C(2), etc. seem to be labeled R1, R2, and so on. R1 and R2 are generic labels for row 1 and row 2 in a coefficient vector. The same setup is used generally for displaying vectors and matrices.
Hint: Unlike series where a number in parentheses indicates a lead or lag, the number following a coefficient is the coefficient number. C(1) is the first element of C, C(2) is the second element of C, and so on.
Hint: Unlike series where a number in parentheses indicates a lead or lag, the number following a coefficient is the coefficient number. C(1) is the first element of C, C(2) is the second element of C, and so on.
Notice that the slope coefficient has been stored in BETA(1), and that since we didn’t reference BETA(2) nothing has been stored in it. Nonlinear Least Squares—343
Notice that the slope coefficient has been stored in BETA(1), and that since we didn’t reference BETA(2) nothing has been stored in it. Nonlinear Least Squares—343
If the estimate runs but doesn’t converge, give the same command again. Since EViews stores the last estimated coefficients in C, the second estimation run picks up exactly where the first one left off. Change iteration limits
If the estimate runs but doesn’t converge, give the same command again. Since EViews stores the last estimated coefficients in C, the second estimation run picks up exactly where the first one left off. Change iteration limits
The basic notion behind GMM is that each of the instruments is orthogonal to a specified function. You can specify the function in any of three ways:
The basic notion behind GMM is that each of the instruments is orthogonal to a specified function. You can specify the function in any of three ways:
Limited Dependent Variables—349
Limited Dependent Variables—349
Dependent Variable: UNION Method: ML - Binary Logit (Quadratic hill climbing) Date: 07/23/09 Time: 16:00 Sample: 11759 Included observations: 1759 Convergence achieved after 6 iterations Covariance matrix computed using second derivative Textbook hint: Textbooks usually describe the relation between probability and inde» in a logit with prob(union= 1)= F(s) rather than prob(union= 1)= 1—F(—s). The two are equivalent for a logit (or a probit), but differ for some other models.
Dependent Variable: UNION Method: ML - Binary Logit (Quadratic hill climbing) Date: 07/23/09 Time: 16:00 Sample: 11759 Included observations: 1759 Convergence achieved after 6 iterations Covariance matrix computed using second derivative Textbook hint: Textbooks usually describe the relation between probability and inde» in a logit with prob(union= 1)= F(s) rather than prob(union= 1)= 1—F(—s). The two are equivalent for a logit (or a probit), but differ for some other models.
Have you tried About EViews on the Help menu and then clicked the button? Only one Nobel prize winner (so far!) appears in the credits list. Which brings us to the topic of autoregressive conditional heteroskedasticity, or ARCH. ARCH, and members of the extended ARCH family, model time-varying variances of the error term. The simplest ARCH model is:
Have you tried About EViews on the Help menu and then clicked the button? Only one Nobel prize winner (so far!) appears in the credits list. Which brings us to the topic of autoregressive conditional heteroskedasticity, or ARCH. ARCH, and members of the extended ARCH family, model time-varying variances of the error term. The simplest ARCH model is:
Hint: Unlike nearly all other EViews estimation procedures, ARCH requires a continu- ous sample. Define an appropriate sample in the Specification tab. If your sample includes a break, EViews will give an error message. In this example, we had to use a subset of our data to accommodate the continuous sample requirement. 352—Chapter 14. A Taste of Advanced Estimation
Hint: Unlike nearly all other EViews estimation procedures, ARCH requires a continu- ous sample. Define an appropriate sample in the Specification tab. If your sample includes a break, EViews will give an error message. In this example, we had to use a subset of our data to accommodate the continuous sample requirement. 352—Chapter 14. A Taste of Advanced Estimation
Dependent Variable: G Method: ML- ARCH (Marquardt) - Normal distribution Date: 12/21/06 Time: 16:08 Sample: 1934M01 1999M10 Included observations: 790 Convergence achieved after 57 iterations Variance backcast used with parameter = 0.7 GARCH = C(16) + C(17)*RESIDG-1)*2 + C(18)*GARCH(C1)
Dependent Variable: G Method: ML- ARCH (Marquardt) - Normal distribution Date: 12/21/06 Time: 16:08 Sample: 1934M01 1999M10 Included observations: 790 Convergence achieved after 57 iterations Variance backcast used with parameter = 0.7 GARCH = C(16) + C(17)*RESIDG-1)*2 + C(18)*GARCH(C1)
Maximum Likelihood—Rolling Your Own Maximum Likelihood—Rolling Your Own—355 Despite EViews’ extensive selection of estimation techniques, sometimes you want to cus- tom craft your own. EViews provides a framework for customized maximum likelihood esti- mation (mle). The division of labor is that you provide a formula defining the contribution an observation makes to the likelihood function, and EViews will produce estimates and the expected set of associated statistics. For an example, we’ll return to the weighted least squares problem which opened the chapter. This time, we’ll estimate the variances and coefficients jointly.
Maximum Likelihood—Rolling Your Own Maximum Likelihood—Rolling Your Own—355 Despite EViews’ extensive selection of estimation techniques, sometimes you want to cus- tom craft your own. EViews provides a framework for customized maximum likelihood esti- mation (mle). The division of labor is that you provide a formula defining the contribution an observation makes to the likelihood function, and EViews will produce estimates and the expected set of associated statistics. For an example, we’ll return to the weighted least squares problem which opened the chapter. This time, we’ll estimate the variances and coefficients jointly.
-ogL: WEIGHTED_EXAMPLE Method: Maximum Likelihood (Marquardt) Date: 12/21/06 Time: 16:28 Sample: 1917M08 2005M04 IF G AND G-1) ncluded observations: 1043 =yaluation order: By observation Convergence achieved after 118 iterations
-ogL: WEIGHTED_EXAMPLE Method: Maximum Likelihood (Marquardt) Date: 12/21/06 Time: 16:28 Sample: 1917M08 2005M04 IF G AND G-1) ncluded observations: 1043 =yaluation order: By observation Convergence achieved after 118 iterations
The estimated coefficients haven’t changed much in this example. The difference between the two esti- mates is that the latter accounts for correlation between the two equations, while the former doesn’t. The Residuals/Correlation Matrix view shows the estimated cross-equation correlation. In this case, there is very little correlation—that’s why the SUR estimates came out about the same the esti- mates from equation-by-equation least squares.
The estimated coefficients haven’t changed much in this example. The difference between the two esti- mates is that the latter accounts for correlation between the two equations, while the former doesn’t. The Residuals/Correlation Matrix view shows the estimated cross-equation correlation. In this case, there is very little correlation—that’s why the SUR estimates came out about the same the esti- mates from equation-by-equation least squares.
Restrictions are linear in coefficients. Vector Autoregressions—VAR
Restrictions are linear in coefficients. Vector Autoregressions—VAR
In Chapter 8, “Forecasting,” we discussed predictions based on ARMA and ARIMA models. This kind of forecasting generalizes, at least in the case of autoregressive models, to multiple dependent variables through the use of vector autoregressions or VARs. While VARs can be quite sophisticated (see the User’s Guide), at its heart a VAR simply takes a list of series and regresses each on its own past values as well as lags of all the other series in the list.
In Chapter 8, “Forecasting,” we discussed predictions based on ARMA and ARIMA models. This kind of forecasting generalizes, at least in the case of autoregressive models, to multiple dependent variables through the use of vector autoregressions or VARs. While VARs can be quite sophisticated (see the User’s Guide), at its heart a VAR simply takes a list of series and regresses each on its own past values as well as lags of all the other series in the list.
Variance decomposition
Variance decomposition
Hint: The dashed lines enclose intervals of plus or minus two standard errors. Vector Autoregressions—VAR—363
Hint: The dashed lines enclose intervals of plus or minus two standard errors. Vector Autoregressions—VAR—363
Forecasting from VARs
Forecasting from VARs
Hint: Since C is a reserved name in EViews, we’ve substituted CONS for C. [o get started with making an EViews model, use Object/New Object... to generate a model named KEYNESCROSS.
Hint: Since C is a reserved name in EViews, we’ve substituted CONS for C. [o get started with making an EViews model, use Object/New Object... to generate a model named KEYNESCROSS.
The model is now complete. Your First Homework—Bam, Taken Up A Notch!—367
The model is now complete. Your First Homework—Bam, Taken Up A Notch!—367
The model window gives details of the solution technique used. Complicated models can be hard for even a computer to solve, but this model is not complex, so the details aren’t very interesting. Note at the bottom of the window that EViews solved the model essentially instantly. Wow hint: The model is nonlinear. There is no closed form solution. This doesn’t bother EViews in the slightest! (Admittedly, some models are harder for a computer to solve.) The model solver created new series containing the solution values. To distinguish these series from our original data, EViews adds “_0” to the end of the name for each solved
The model window gives details of the solution technique used. Complicated models can be hard for even a computer to solve, but this model is not complex, so the details aren’t very interesting. Note at the bottom of the window that EViews solved the model essentially instantly. Wow hint: The model is nonlinear. There is no closed form solution. This doesn’t bother EViews in the slightest! (Admittedly, some models are harder for a computer to solve.) The model solver created new series containing the solution values. To distinguish these series from our original data, EViews adds “_0” to the end of the name for each solved
Looking At Model Solutions
Looking At Model Solutions
Looking At Model Solutions—371
Looking At Model Solutions—371
Teminology hint: Exogenous variables are determined outside the model and their val- ues are not affected by the model’s solution. Endogenous variables are determined by the solution of the model. Think of exogenous variables as model inputs and endoge- nous variables as model outputs.
Teminology hint: Exogenous variables are determined outside the model and their val- ues are not affected by the model’s solution. Endogenous variables are determined by the solution of the model. Think of exogenous variables as model inputs and endoge- nous variables as model outputs.
Hint: The estimated standard error is used when you ask EViews to execute a stochas- tic simulation, a feature we won’t explore, referring you instead to the User’s Guide. Hint: How come the equation shows only CONS + 1+ G? What happened to exports and imports—not to mention that discrepancy thing? In the equation view, EViews displays only enough of each equation so that you can remember which equation’s which. To see the full equation, switch to the text view or look at the equation’s Prop: erties.
Hint: The estimated standard error is used when you ask EViews to execute a stochas- tic simulation, a feature we won’t explore, referring you instead to the User’s Guide. Hint: How come the equation shows only CONS + 1+ G? What happened to exports and imports—not to mention that discrepancy thing? In the equation view, EViews displays only enough of each equation so that you can remember which equation’s which. To see the full equation, switch to the text view or look at the equation’s Prop: erties.
Vanity hint: In the problem at hand, all we’re doing is making the answer look pretty. In more complicated problems a smaller convergence limit has the advantage that it helps assure that the computer reaches the right answer. The disadvantages are that the solution takes longer, and that sometimes if you ask for extreme accuracy no satis factory answer can be found.
Vanity hint: In the problem at hand, all we’re doing is making the answer look pretty. In more complicated problems a smaller convergence limit has the advantage that it helps assure that the computer reaches the right answer. The disadvantages are that the solution takes longer, and that sometimes if you ask for extreme accuracy no satis factory answer can be found.
Hint: We chose the name “G_1” because the suffix has to match the scenario alias.
Hint: We chose the name “G_1” because the suffix has to match the scenario alias.
Hint: You can override an exogenous variable but you cannot override an endogenous variable because the latter would require a change to the structure of the model.
Hint: You can override an exogenous variable but you cannot override an endogenous variable because the latter would require a change to the structure of the model.
Your Second Homework—377
Your Second Homework—377
Models can be used for solving complicated systems of equations under different scenarios, as we’ve done above. Models can also be used for forecasting dynamic systems of equa- tions. This is especially useful in forecasting from vector autoregressions.
Models can be used for solving complicated systems of equations under different scenarios, as we’ve done above. Models can also be used for forecasting dynamic systems of equa- tions. This is especially useful in forecasting from vector autoregressions.
Hint: If you prefer, instead of creating a model and then copying in a VAR you can us Proc/Make Model from inside the VAR to do both at once. The model taught in our introductory economics course was linear because it’s hard for peo- ple to solve nonlinear models. Computers are generally fine with nonlinear models,
Hint: If you prefer, instead of creating a model and then copying in a VAR you can us Proc/Make Model from inside the VAR to do both at once. The model taught in our introductory economics course was linear because it’s hard for peo- ple to solve nonlinear models. Computers are generally fine with nonlinear models,
Documentary Hint: It’s sometimes important to keep a record of your analysis. While you can print lines typed in the command pane, or save them to disk as “COM- MAND.log”, point-and-click operations aren’t recorded for posterity. Entering com- mands in a program and then running the program is the best way to create an audit trail in EViews. Hint: If you choose the Quiet radio button in the Run Program dialog (not shown here; see below), programs execute faster and more peacefully. But Verbose is some- times helpful in debugging.
Documentary Hint: It’s sometimes important to keep a record of your analysis. While you can print lines typed in the command pane, or save them to disk as “COM- MAND.log”, point-and-click operations aren’t recorded for posterity. Entering com- mands in a program and then running the program is the best way to create an audit trail in EViews. Hint: If you choose the Quiet radio button in the Run Program dialog (not shown here; see below), programs execute faster and more peacefully. But Verbose is some- times helpful in debugging.
Program Variables
Program Variables
If the step value is omitted, EViews steps the control variable by 1.
If the step value is omitted, EViews steps the control variable by 1.
An EViews program is essentially a list of commands available to execute as needed. You can make the commands apply to different objects by providing arguments when you run the program. You can also automate repetitive commands by using numerical and string “for...next” loops. An EViews program is an excellent way to document your operations, and compared to manually typing every command can save a heck of a lot of time!
An EViews program is essentially a list of commands available to execute as needed. You can make the commands apply to different objects by providing arguments when you run the program. You can also automate repetitive commands by using numerical and string “for...next” loops. An EViews program is an excellent way to document your operations, and compared to manually typing every command can save a heck of a lot of time!
Hint: Any untitled EViews object will disappear when you close its window. If you want to keep something you’ve frozen, use the button.
Hint: Any untitled EViews object will disappear when you close its window. If you want to keep something you’ve frozen, use the button.
aneq. output The Command and Programming Reference includes extensive tables documenting each object type and the commands and views associated with each.
aneq. output The Command and Programming Reference includes extensive tables documenting each object type and the commands and views associated with each.
QMS also posts free, minor enhancements from time to time and, once-in-a-while, posts updated documentation. On occasion, QMS releases a minor version upgrade (5.1 from 5.0, for example) free-of-charge to current users. Some of these “minor” upgrades include quite significant new features. It pays to check the EViews website.
QMS also posts free, minor enhancements from time to time and, once-in-a-while, posts updated documentation. On occasion, QMS releases a minor version upgrade (5.1 from 5.0, for example) free-of-charge to current users. Some of these “minor” upgrades include quite significant new features. It pays to check the EViews website.
Date Representation
Date Representation
Option-al Recommendations—405
Option-al Recommendations—405
Spreadsheet Defaults display group. This allows you to set the numeric display characteristics individually for each type of data in the list. To switch all your data to Comma as decimal, you’ll want to switch to each item in the Data displayed as drop-down and select Comma as decimal. You do not need to click OK and exit the dialog between each setting; EViews remembers each group of settings individually until you click OK.
Spreadsheet Defaults display group. This allows you to set the numeric display characteristics individually for each type of data in the list. To switch all your data to Comma as decimal, you’ll want to switch to each item in the Data displayed as drop-down and select Comma as decimal. You do not need to click OK and exit the dialog between each setting; EViews remembers each group of settings individually until you click OK.
Keyboard Focus The radio button Keyboard focus directs whether typed characters are “sent” by default to the command pane or to the currently active window. Most people leave this one alone, but if you find that you’re persistently typing in the command pane when you meant to be edit- ing another window, try switching this button and see if the results are more in line with what your fingers intended.
Keyboard Focus The radio button Keyboard focus directs whether typed characters are “sent” by default to the command pane or to the currently active window. Most people leave this one alone, but if you find that you’re persistently typing in the command pane when you meant to be edit- ing another window, try switching this button and see if the results are more in line with what your fingers intended.
Page After Page” for an extended discussion of frequency conversion. Font Options
Page After Page” for an extended discussion of frequency conversion. Font Options
Internally, each observation in a series takes up eight bytes of storage. There’s no great rea- son you should care about this as either you have enough memory—in which case it doesn’t matter—or you don’t have enough memory—in which case your only option is to buy some more. And as a practical matter, unless you’re using millions of data points the issue will never arise.
Internally, each observation in a series takes up eight bytes of storage. There’s no great rea- son you should care about this as either you have enough memory—in which case it doesn’t matter—or you don’t have enough memory—in which case your only option is to buy some more. And as a practical matter, unless you’re using millions of data points the issue will never arise.
Hint: There’s no reason not to use compression, so use it, unless someone using a ver- sion of EViews earlier than 5.0 needs to read the file, in which case—don't.
Hint: There’s no reason not to use compression, so use it, unless someone using a ver- sion of EViews earlier than 5.0 needs to read the file, in which case—don't.
Graphics Defaults
Graphics Defaults
EViews automatically names its options-storing-file “EViews32.ini”. To store multiple ver- sions of “EViews32.ini”, fine tune your options to suit a particular purpose, then reset the Ini File Path to a unique path for each version you wish to save. There’s no browsing for “EViews32.ini”, and the name “EViews32.ini” is hard-coded into the program, so to use multiple option sets you need to remember the paths in which you’ve stored each set.
EViews automatically names its options-storing-file “EViews32.ini”. To store multiple ver- sions of “EViews32.ini”, fine tune your options to suit a particular purpose, then reset the Ini File Path to a unique path for each version you wish to save. There’s no browsing for “EViews32.ini”, and the name “EViews32.ini” is hard-coded into the program, so to use multiple option sets you need to remember the paths in which you’ve stored each set.

Related papers

EViews 7 Users Guide II
Nesrin Eskiocak Zubaroglu

Part VIII. "Panel and Pooled Data" on page 563 documents working with and estimating models with time series, cross-sectional data. The analysis may involve small numbers of cross-sections, with series for each cross-section variable (pooled data) or large numbers systems of cross-sections, with stacked data (panel data).

View PDFchevron_right
CAUSALITAS GRANGER ANALYSIS.docx
Muhamad Armawaddin

The classical flow methodology, however, changed in 1983, pioneered by Hendry and Richard, which is either% u201Ctop down or general to specific% u201D.

View PDFchevron_right
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved