Tutorial7 Dummy Variables

8/10/2019 Tutorial7 Dummy Variables

1/47

EViews TrainingDummy Variables

Note: Data and workfiles for this tutorial are provided in:

Data: Data.xlsx

Results: Results.wf1

Practice Workfile: Data.wf1


2/47

Data and Workfile Documentation

Data.wf1and Data.xlsx consist of two pages (tabs) with the following data:

Workfile Page: Dated(Data.xlsx tab Dated): daily, from Jan 1, 1950August 2, 2012

VolumeS&P 500 volume (source: St. Louis Federal Reserve Bank, FRED Database)

Volume1- percent change in S&P500 Volume

ReturnS&P 500 return (source: St. Louis Federal Reserve Bank, FRED Database)

Workfile Page: Cross_Sect ion (Data.xlsx tab Cross_Section) data on 526 individuals

wagewage per hour (in dollars)

educnumber of years of education

experexperience, number of years

femaledummy variable, equal to 1 if female, 0 otherwise

marr ieddummy variable, equal to 1 if married, 0 otherwise

2

* These data are from Wooldridge, Introductory Econometrics(4thEdition).


3/47

Data Series Objects: Series and Groups

It is fairly easy to create dummy variables in EViews by using sample

commandsor a number of EViews functions:

smp l

@recode

@date

@expand

This tutorial demonstrates how to create:

Dummy Variables using samples

Dummy Variables using the @recodefunction

Date Dummy Variables using the @datefunction

Categorical Dummy Variables using the @expandfunction

3


4/47

Creating Dummy VariablesUsing Samples


5/47

Creating Dummy Variables Using Samples:

Example 1

The easiest way to create dummy variables in EViews is by using samples

(smp lcommand).

Lets illustrate a few examples using the Datedpage in Work fi le Data.wf 1.

Suppose you would like to create a dummy variable equal to 1, if return>0.2,

and 0 otherwise.

5

Dummies with samples: Example 1

1. Open Data.wf1workfile. Click on the

Datedpage.

2. Type in the command window:

smpl if return>0.2

series dummy1=1

smpl if return


6/47


Example 1 (contd)

The original series (return) and the new dummy variable are both shown here.

Original series Dummy Variable

6


7/47


Example 2

7

Now suppose you would like to create a dummy variable equal to 1, if

return>0.3, or return0.3 or return-0.3 and return


8/47

Creating Dummy VariablesUsing @recode function


9/47

Creating Dummy Variables Using @recode

Another way to create dummy variables in EViews is by using @recode

function.

9

Function Description

@recode(s,x,y) Returns x if condition s is true; otherwise it returns y


10/47

Creating Dummy Variables Using @recode:

Example 1

10

Suppose you want to create a dummy equal to 1 ifreturn


11/47


Example 1 (contd)

The original series (return) and the new dummy variable are both shown here.

Original series Dummy Variable

11


12/47


Example 2

12

Create a dummy variable equal to 1 if -0.2


13/47


Example 3

13

Create a dummy variable equal to 1 if return=1, 0 otherwise.

Dummies with @recode: Example 3

1. Click on Dated workfile page. Type in the command window:

series dummy6=@recode(return=1,1,0)

2. Press Enter.

Original Series Dummy Variable


14/47


Example 4

14

Suppose you would like to create two new series by separating an

existing series (return) as follows;

series1 collects values of returnfor which -0.3


15/47


Example 4 (contd)

15

The original series (return) and the two new dummy variables are shown here.

Original series Series 1 Series 2


16/47


Example 5

16

Lets create a total return index from daily S&P500 returns.



series index=@recode(@trend=0,100, index(-1)*(1+return/100))

2. Press Enter.

Original Series Index


17/47


Example 6

17

Lastly, suppose you want to create a series that excludes outliers from the

returnseries in the previous example. Specifically, suppose that the new series excludes the highest and the lowest

2.5% values of returns.



series no_outlier=@recode(return>@quantile(return,0.025) and

return


18/47


Example 6 (contd)

18

The original series (return) and the new series are shown here.

Original series No Outlier


19/47

A Few Notes on Simple Dummy Variables

For simple dummies, you dont need to use smp lor @recode.

You can create them simply by defining the logical expression directly in the

command window, as shown in the following examples.

However, you may have to use smp lor @recode for more complex cases.

19


20/47

A few Notes on Simple Dummy Variables:

Example 1

20

Create a dummy equal to 1 if return


21/47

A few Notes on Simple Dummy Variables:

Example 2

21

Create a dummy equal to 1 if return=0.1, and 0 otherwise.

Simple Dummies: Example 2

1. Click on Dated workfile page. Type in the command

window:

series d2=return=0.1

2. Press Enter. Original Series Dummy Variable


22/47

Date Dummy Variables


23/47

Creating Date Dummies

Dated dummies can be created by using @recode and @dateor @dateval.

For more details on date functions, see tutorial on Date Functions.

23

Funct ion Description

@date Returns the date associated with each observation

@dateval Returns the date associated with a text representation of a date

@year Returns the year in which each observation begins

@quarter Returns the quarter of the year in which each observation begins

@month Returns the month of the year in which each observation begins

@day Returns the day of the month in which each observations begins

@weekday Returns the day of the week

Main Functions


24/47

Creating Date Dummies:

Example 1

24

Suppose you want to create a dummy variable equal to 1 for all dates after

1995/03/15 and 0 otherwise.Date Dummies: Example 1

1. Click on the Datedworkfile page. Now, type in the command window:

series dumdate1=@recode(@date>@dateval("1995/3/15"),1,0)

2. Press Enter.

Alternatively, without @recode:

1. Type in the command window:series dumdate1=@date>@dateval("1995/3/15")

2. Press Enter.

Dummy Variable


25/47


Example 2

25

Create a dummy variable equal to 1 for all dates before April 1979 and after

February 1994.Date Dummies: Example 2

1. Click on the Datedworkfile page. Type in the command window:

series dumdate2=@recode(@date@dateval("1994m2"),1,0)

2. Press Enter (please type the command in one line).

Alternatively, without @recode:


series dumdate2=@date@dateval("1994m2")

2. Press Enter(please type the command in one line).


26/47

26


Example 2 (contd)

Dummy Variable

A graph of the new dummy variable is shown here.

As you can see, it assumes a value of 1 prior to April 1979 and postFebruary 1994.


27/47


Example 4

27

Dated dummies can also be created using @year, @month, @day date functions.

Create a dummy equal to 1 if the month is January or the day of the week is Friday.

Date Dummies: Example 4


series dumdate4=@recode(@month=1 or @weekday=5,1,0)

2. Press Enter.

Alternatively, without @recode:1. Type in the command window:

series dumdate4=@month=1 or @weekday=5

2. Press Enter.

Dummy Variable


28/47


Example 5

28

Create a dummy equal to 1 for all years after 1994 if the month is January or the

day of the week is Friday.

Date Dummies: Example 5


series dumdate5=@recode(@year>1994 and (@month=1 or @weekday=5),1,0)

2. Press Enter.


series dumdate5=@year>1994 and (@month=1 or @weekday=5)

2. Press Enter.


29/47


Example 6

29

Create a dummy variable equal to 1 for all months when there is a presidential

election.Date Dummies: Example 6


smpl @allseries dumpres=0

smpl if @mod(@year,4)=0 and @month=11

series dumpres=1smpl @all

2. Press Enter after each command line.

Dummy Variable


30/47

Dummy Variables in Non-DatedWorkfiles


31/47

Dummies in Non-dated Workfiles

You can follow the methodology used in Dated Dummies to create dummies

in non-dated workfiles.

Simply use @obsnum (instead of @date) to create the dummy variables.

31


32/47

Creating Dummies in Non-Dated Workfiles:

Example 1

32

In the Cross-Sect ionpage, create a dummy variable equal to 1 after the 5th

observation, and 0 otherwise.Dummies in non-Dated Workfiles: Example 1

1. Click on the Cross-Sect ionworkfile page. Now, type in the command window:

series dum1=@recode(@obsnum>5,1,0)

2. Press Enter.

Dummy Variable


series dum1=@obsnum>5

2. Press Enter.


33/47

Creating Dummies in Non-Dated Workfiles:

Example 2

33

In the Cross-Sect ionpage, create a dummy variable equal to 1 for all the

observations between the 7th

and 12th

observationDummies in non-Dated Workfiles: Example 2

1. Click on the Cross-Sect ionworkfile page. Now, type in the command window:

series dum2=@recode((@obsnum>=7 and @obsnum=7 and @ obsnum


34/47

Creating Dummy VariablesUsing@expand function


35/47

Creating Categorical Dummies

Categorical dummies can be easily created by using @expand function.

35

Function Description

@expand Allows you to create a group of dummy variables by expanding

out one or more series into individual categories


36/47

Creating Dummy Variables using @expand:

Example 1: Categorical Dummies

Lets create a group of series as follows:

One series with 1 if female and 0 if male

One series with 1 if male and 0 if female

36

Dummies using @expand: Example 1

1. Click on the Cross-Sect ionworkfile page. Type in the command window:

group g1 @expand(female)

2. Press Enter.

The group is shown here withthe two series separating

males from females.


37/47



Now lets create a new group of series as follows:

One series containing 1 if male and single. One series containing 1 if male and married.

One series containing 1 if female and single.

One series containing 1 if female and married.

37


1. Click on the Cross-Sect ionworkfile page. Type in the command window:group g2 @expand(female, married)

2. Press Enter.


38/47



Now lets create a new group of series as follows:

One series equal to educ(years of education) if male (female=0). One series equal to educif female (female =1).

38


1. Click on the Cross-Sect ionworkfile page. Type in the command window:

group g3 educ*@expand(female)

2. Press Enter.

The group is shown here with the twoseries showing the years of

education for males separately from

years of education for females.

C @


39/47

Dummies can also be created by using @expand in conjunction with other

date functions (@year, @month, @day, etc.).

For example, suppose you would like to create a dummy variable for each

day of the week.


Example 4: Date Dummies

39



group g1 @expand(@weekday)

2. Press Enter.

C i D V i bl i @ d


40/47

Suppose that now you would like to create a dummy variable for each year.


Example 5: Date Dummies

40



group g2 @expand(@year)

2. Press Enter.


41/47

Dummy Variables in Regressions

D i i R i


42/47

Dummies in Regressions:

Example 1

In EViews you can use dummy variable expressions in regressions without

having to first create and save the dummies. Suppose you would like to estimate a regression of returnon vo lume1and a

dummy variable equal to 1 for all dates after 1994/12/2.

42

Dummies in Regressions: Example 1

1. Click on the Datedworkfile page. Type in the

command window:

equation eq1.ls return c volume1

@date>@dateval("1994/12/2")

2. Press Enter (type the command in one line).

Alternatively:

1. Select QuickEstimate Equation from

the top menu bar.

2. The Equation Estimation box opens up.

Specify here your equation (as shown in

the figure)

3. Click OK.

D i i R i


43/47


Example 1 (contd)

43

The estimation output is shown here.

As you can see EViews estimates the coefficient of the dummy variable

directly without us having to create and save the date dummy first.

*Note: see tutorial on Basic Estimation for details

on regression analysis in EViews.

D i i R i


44/47


Example 2 Now suppose you would like to estimate a regression of returnon vo lume1and

a dummy variable equal to 1 for all Januaries since 1994/12/2.

44



equation eq2.ls return c volume1 (@date>@dateval("1994/12/2") and @month=1)

2. Press Enter.

Note that there are no spaces in

the logical expression:

@month=1 is correct @month = 1 is not correct

D i i R i


45/47


Example 3

45

Now lets carry out a regression in the Cross_Sect ion workfile page.

Suppose we want to regress wageon exper, educand four dummy variablescreated by @expand(female,married).


1. Click on the Cross_Sect ionworkfile page.

Type in the command window:

equation eq1.ls wage exper educ

@expand(female, married)

2. Press Enter (type command in one line).

D i i R i


46/47


Example 3 (contd)

46

In the previous example, notice that the constant was not included.

If you include a constant, then EViews will be unable to estimate the

regression and return an error message because of perfect collinearity (the

dummy variable trap).

You will receive an error message

(shown here). In order to estimate the regression with

a constant, you should exclude one of

the dummies.



equation eq1.ls wage c exper educ@expand(female, married)


D i i R i


47/47

47

Lets exclude one of the dummies. You can use either one of the following

commands: @dropf i rs t (drops the first dummy).

@droplast (drops the last dummy).


Example 3 (contd)



equation eq2.ls wage c exper educ@expand(female, married,@dropfirst)


Date post:	02-Jun-2018
Category:	Documents
Upload:	rosa-maria-vega-torres
View:	238 times
Download:	0 times

Tutorial7 Dummy Variables

Documents