[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: gologit2

From   Richard Williams <>
Subject   RE: st: gologit2
Date   Thu, 17 Apr 2008 22:29:47 -0500

At 03:15 PM 4/17/2008, Maarten buis wrote

Alternatives are:
o If your dataset is large you can use (Raftery 1995), he does not deal
with -ologit- specificaly, but more generally with hypothesis
testing/model selection in large datasets. Than you can compare BICs,
these are less likely to reject the proporitonal odds assumption, and
if they do show strong evidence against the proportional odds
assumption you should probably be worried anyhow.
I agree.

o You can use other methods for dealing with ordinal dependent
variables that are not surrounded by these ingrained practices, for
instance the stereotyped ordered logit (-slogit- in Stata). This is a
cheat, but if it gets you around the referees...
Of course, if you want to cheat, you could just run oprobit, and hope the referees don't realize that oprobit models can have the same problems as ologit models. You just use more general terminology such as parallel lines or parallel regressions rather than proportional odds. I actually had an economist tell me once that oprobit was better than ologit because only ologit had a problem with proportional odds. I thought this was kind of silly, because proportional odds just happens to be a consequence of parallel lines when using the logit link. Regardless of whether you are using ologit or oprobit, the method assumes parallel lines.

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

* For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index