Heteroskedasticity test for logit and logistic models

[andrea2000]

~~

I am conducting research which involves a logistic and logit model. I have a binary dependent variable and several independent variables. If I wanted to test for heteroskedasticity in Stata, is there a command for this?

[Maarten Buis]

That question was answered here: http://www.statalist.org/forums/foru...roscedasticity

[andrea2000]

Thanks Maarten. So for logistic/logit and probit models, do I need to worry about heteroskedasticity?

[Maarten Buis]

Yes and no. Yes, it is an issue, as described in the article of Rich I linked to. No, because there is nothing you can do (attempts exist, but they are just too fragile to be of practical use).

[Joao Santos Silva]

There may be a difference of cultures here, but some economists worry about and test for heteroskedasticity in binary choice models. Davidson and MacKinnon have a seminal paper on this:

Davidson, R. and MacKinnon, J.G. (1984), "Convenient specification tests for logit and probit models", Journal of Econometrics, 25, 241-262.

For example, the RESET test can be interpreted as a test for heteroskedasticity, which shows the link between heteroskedasticity and functional form in this context.

Joao

[Richard Williams]

The Davidson-MacKinnon test is programmed in Shazam:

http://shazam.econ.ubc.ca/intro/logit3.htm

So, if really really desperate, either get Shazam, or figure out how to adapt the Shazam code to Stata.

My theory is that the closest thing to a Stata equivalent would be something like

hetprob y x1-x10, het(x1-x10)

Sometimes that model converges quickly, sometimes it never converges at all. You would just have to see what happens with your data.

Here is a blogger who is big on doing a hetero test with logit:

http://davegiles.blogspot.com/2011/05/gripe-of-day.html

But here is a poster who raises concerns about it:

http://www.carlislerainey.com/2013/0...probit-models/

[Tigistu Tgew]

hello, i'm conducting research which involves PSM model (impact studies), how can i choose among matching algorithms? if any stata command?

[Tigistu Tgew]

hello, i'm conducting research which involves PSM model (impact studies), how can i choose among matching algorithm? if any stata command?

[Richard Williams]

Welcome to Statalist.

i recommend starting a new thread with a clear title rather than tacking on to a very old and unrelated thread. The people who know the answer may never look at a seemingly unrelated thread.