Fixed effects panel regression (LSDVM)

[chad]

Hi there,

I am a new Eviews user and I am trying to do a simple Least Squares Dummy Variable model. I have attached an example excel: Y is the dependent variable, X1, X2 and X3 are the independent variables. You can also see the dates and "Country" is supposed to define the cross section for three countries. Could you tell me if my data is in correct form in excel to do LSDVM pls?

Thanks

[startz]

That should work.
Read the file into EViews and see what happens.

[chad]

It didnt recongnise the country identifiers as texts for some reason, I changed them to numbers (instead of "country1", just "1"; instead of "country2", just "2" etc) and it worked. As I say, I am a very new user of eviews and I want to make sure I am selecting the correct options when doing LSDVM. I did the following in my workfile building process and equation building process. Could you pls confirm it is correct?

New Workfile: Structure type "balanced panel" selected; # of cross sections: 4

Then: Proc/Import/Import file and selected the excel file; defined names cross section series and date series. Basic structure selected: "Dated Panel".

In Equation setting: Method selected "LS - Least Squares (LS and AR); Panel Options: Cross-section "fixed", Period "none", GLS Weights "no weights", Coef covariance method "Ordinary".

my excel and workfile are attached. Can you pls confirm pls?

Thanks

[startz]

It looks correct to me. But unless this is just made up data something is wrong since you are getting a perfect fit.

[chad]

yes this is made up data, so no worries about the perfect fit.

I have another question. If I want to make just the contant vary across cross section but other coefficiensts remain the same. What change do I need to make in the code (also maybe in excel file) pls?

I was following from other forum discussion (http://forums.eviews.com/viewtopic.php?f=4&t=1087), in which your colleague Eviews_Gareth suggests using "@expand(@crossid)" command.

However if I keep all the settings same and change my equation from "y c x1 x2 x3" to "y @expand(@crossid) X1 x2 x3" to achieve this, it gives an error message of "near singular matrix".

Can please advice what I need to do here to fix this issue in order to get varying constants but same coefficients across cross section.

Thanks again..

[startz]

If you follow Gareth's suggestion with real data it will work fine.

[chad]

Okay i will try that.. What adjustment i need to make in order avoid 'dummy variable trap' in my LSDVM pls? Or Eviews inherentlt takes care of this when i define my data set as in the attached excel above?

Thanks

[startz]

If you use @expand, just leave out the constant "C".

[chad]

Sorry i am a little confused. Are you suggesting (assuming i am using real data):

If i use in my equation 'y @expand(@crossid) x1 x2 x3' instead of 'y x1 x2 x3' with the data exactly in same format in excel (ie no new dummy column definitions manually or anything) and same panel options as before, then i will get different constants, but same coefficients across cross sections and at the same time avoid the 'dummy variable trap'' in the model?

Thanks

[startz]

Exactly.