Statalist


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

Re: st: marginal effect of clogit


From   "Clive Nicholas" <clivelists@googlemail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: marginal effect of clogit
Date   Sun, 23 Sep 2007 01:58:33 +0100

Somsupa Nopprach replied:

[...]

>   I notice that the significance of the explanatory
> variable in the xtlogit and in oglm are different.
>
>   I means that eventhough in -xtlogit- model the first
> variable, for example, is not statistically
> significant while it does significant in -oglm-
> command. So how i can interpret this ? Should i said
> that the first variable really have a significant
> effect on the dependent variable?

Why do you expect the models to be the same? Would you expect fitting
a linear probability model with -reg- and then by -glm- to be the
same? No, because they are _two different routines_, and so are these
two. The latter model gives a better fit to the data than the former.
If the assumptions underpinning -oglm- have not undermined in the
process, that should be a cause for celebration. You have to take the
statistical output of each model on their merits. The interpretation
of the parameters on the variables should not change. But, as always,
I stand to be corrected.

>   Secondly, could you explain me more why you said
> -oglm- is useful eventhough -oglm- does not fix the
> problem.

It is useful because, as I tried to show you, you get an idea of how
inappropriate the use of an interaction term (or terms) is for your
logit model, if at all. If Allison's delta is very small, then one may
conclude that you could compare your logit coefficient(s) across your
groups of interest legitimately. If Allison's delta is high, however,
you have a problem, as well as a decision to make: do you press on
regardless, or do you bite the bullet and forego the interaction
terms? One point I failed to emphasise last time out was that by
multiplying delta by 100, you obtain the percentage difference in the
disturbance variance across groups.

It is also worth saying that Allison's delta is not the only method
that estimates this. If you dig around Richard Williams' class papers
at http://www.nd.edu/~rwilliam/, then you should find a paper in which
he discusses, compares and critiques all of these methods, including
Allison's.

-- 
Clive Nicholas

[Please DO NOT mail me personally here, but at
<clivenicholas@hotmail.com>. Thanks!]

"Courage is going from failure to failure without losing enthusiasm."
-- Winston Churchill
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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   |   What's new   |   Site index