Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: oglm and heterogeneous choice models


From   Richard Williams <richardwilliams.ndu@gmail.com>
To   statalist@hsphsun2.harvard.edu, statalist@hsphsun2.harvard.edu
Subject   Re: st: oglm and heterogeneous choice models
Date   Tue, 25 Oct 2011 17:00:18 -0500

At 02:55 PM 10/25/2011, Rourke O'Brien wrote:
Gotcha. That makes sense. I was trying to find another way to diagnose
which variables predict residual variance since using "sw, pe(.05) lr:
oglm y x1 x2 x3 x4 , eq2(x1 x2 x3 x4) flip" wont converge unless I
reduce the model and use fewer predictors.

The gologit2 results do indeed mirror the logit results. I was just
wondering if I could use gologit2 to see what variables would be
problematic.

Not with a dichotomous dv. And given that it is a dichotomous DV you could just use hetprob, which is probably faster than oglm and may be better at catching esoteric problems.

I am interested in testing an interaction between sex and income in
predicting success on a dichotomous variable. If i model using oglm
without the interaction and ", het(sex)" the lnsigma for sex is
significant. When I include the sexXincome interaction with ",
het(sex)" the lnsigma for sex is no longer significant (or even
close). I interpret this to mean the interaction has in effect dealt
with the hetero problem, correct?

Hopefully. As I note in my SMR piece (full citation is in the oglm help), sometimes you can have an interaction term or a hetero term, but not both at the same time. This might be because of collinearity between the terms. Overlap between vars in the choice and variance equations will be even more of a problem with a dichotomous dv. My inclination would be to declare victory at this point with the interaction model.

But how can I tell if other predictors in the model that might also be
problematic?

You probably just have to rely on what makes sense to you. There are always zillions of things that could be tossed into an equation, e.g. more variables, interaction terms, squared terms. You can't test everything but you can look at the more theoretically plausible things.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index