[Camp Assignment 3 Question] Regression with multiple dummy variables

[Seungmin Lee]

Gedeon and I are working on the assignment 3 which is mainly regression.

Table 2 requires regression model controlling caste group, religion and state, which are all multinomial variables.

Gadeon and I made regression with fixed effects by creating multiple variables (ex. reg ... i.state i.caste i.religion ...),
and we have regression results for each state, caste and religion.

Is it OK to have all those results in our table?, If not, is there any way to regress them without creating multiple dummy variables?
(we once tried to regress with within-transformation (xtset fe), but it does not have timely data...)


* One more short question: should we allow heteroskedasticity-standard error by including 'robust'?

[Ajay Shenoy]

Check out the drop() option on esttab (or alternatively, the keep option if that's easier).

And yes, use the robust standard errors.  They will probably be almost identical, but technically speaking there's no reason to expect homoskedasticity in any of these variables.

-@