Statalist


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

st: clogit vs. logit


From   "Regina Kemmler" <Regina.Kemmler@gmx.de>
To   statalist@hsphsun2.harvard.edu
Subject   st: clogit vs. logit
Date   Tue, 10 Nov 2009 09:56:43 +0100

Dear Statalist,

I'd like to conduct a clogit analysis.
I have a sample of several household for whom I'd like to calculate the beta coefficient for 4 different models and conduct the likelihood ratio test afterwards to test which model fits best. Unfortunately the clogit doesn't work for all models. But as the clogit and logit are supposed to give the same results given you have one group I did the following syntax.

keep if (id==1)
clogit IV DV1 DV2, group(id)
gen b1_DV1=_b[DV1]
gen b1_DV2=_b[DV2]
gen lik1=e(ll) 
estimates store m1

logit IV DV2
gen b2_DV2=_b[DV2]
gen kons2=_cons
gen lik2=e(ll)
estimates store m2

logit IV DV1
gen b3_DV=_b[DV]
gen kons3=_cons
gen lik3=e(ll)
estimates store m3

logit IV
gen kons4=_cons
gen lik4=e(ll)
estimates store m4

lrtest m1 (m2 m3 m4), stats


Is it okay to mix up clogit and logit? Does it really give the same results? Why is it then not possible to calculate clogit on the latter 3 models? And: is it possible to conduct a likelihood ratio test if I mix up clogit and logit?

Your help is much appreciated!
Thanks a lot,
Regina



-- 
Jetzt kostenlos herunterladen: Internet Explorer 8 und Mozilla Firefox 3.5 -
sicherer, schneller und einfacher! http://portal.gmx.net/de/go/chbrowser
*
*   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   |   What's new   |   Site index