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

Re: st: gologit2

From   Maarten buis <>
Subject   Re: st: gologit2
Date   Tue, 15 Apr 2008 17:30:31 +0100 (BST)

--- Christian Bustamante <> wrote:
> If the problem is in only a small subset of variables, it could be
> reasonably use gologit? (I suppose, that you can find these variables
> with the -brant, detail- command.
-gologit2- is designed for the situation where the proportional odds
assumption is violated for some but not all variables. Having said
that, it is up to you to determine whether you believe the Brant test
(I am currently running a simulation on that, and will report the
results when they are in) and to determine that -gologit2- is otherwise
appropriate. This latter point can only be answered by the researcher
who has access to the data, knows how the data was collected, what the
research question is, what the theory is, etc. In the end you are
responsible for your analysis, you cannot delegate that responsibility
to some test, the statalist, or some eminent statistician /

> When I did the Brant test, some chi-squares values of variables
> appears as negatives (and p-value equal to 1), how can in be
> possible?

The Brant test performs a series of -logit-s, so it does not respect
the fact that each individual belongs to one and only one category.
This can sometimes lead to weird results. 
> Which representative author has written about this estimation
> technique?

Look at the references in:

Hope this helps,

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Yahoo! For Good helps you make a difference
*   For searches and help try:

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