How to add county fixed effect on a pooled cross-sectional data?

[Alex Carr]

Hello Statalists,

I have a pooled cross-sectional data at individual house level i for the whole state of California. I want to add the county fixed effect for each county j to capture some county specifics. How do I do this?

I tried to add

 i.counties

to the regression but it says "actor variables may not contain noninteger values". Any clues? Thanks


Alex

[Marcos Almeida]

It seems the "counties" variable is not a categorical one. You may check this out by typing descriptive commands, before fixing it.

[Alex Carr]

It seems the "counties" variable is not a categorical one. You may check this out by typing descriptive commands, before fixing it.

Hi, i did this by typing

tabulate county, gen dcounty

and it generated about 800 dummies (dcounty1, dcounty2...dcounty 800) but when I tried to add

i.dcounty

to my regression it says dcounty ambiguous abbreviation. any clues?

[Rocio Fernandez Mendez]

Hi,
I don't think you need to create those dummies before hand.
If county is a categorical variable, I believe you should be able to use i.county directly, and the reg command generates the dummies for you.

[Carole J. Wilson]

Rocio is correct. Your regression statement should include i.county, not i.dcounty.

[Alex Carr]

Hi,
I don't think you need to create those dummies before hand.
If county is a categorical variable, I believe you should be able to use i.county directly, and the reg command generates the dummies for you.

what do you mean by categorical variable. I have like 800 different counties. does that mean i should have 800 indicators?

[Carole J. Wilson]

Using the following will produce a regression with coefficients for each county (minus one used as the base category) plus your other independent variables—so 799 county coefficients:

reg yvar x1 x2 x3 i.county