Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: gllamm (was st: ologit - conditional?)

From   Joseph Coveney <[email protected]>
To   Statalist <[email protected]>
Subject   Re: gllamm (was st: ologit - conditional?)
Date   Sun, 24 Jul 2005 02:51:27 +0900

Ricardo Ovaldia wrote:

Thank you again. One more question please. I have
another outcome in this experiment which is a
percentage ranging from 0 to 50, would it make sense
to use -gllamm- with a Gaussian family and an identity
link. Or is there a better way.


If you want to fit normal mixed models, then -xtreg- or -xtmixed- would be 
better choices:  -xi: xtreg percentage i.culture_medium, i(dam)-.  The 
untransformed response variable might not be adequately normal-like for this, 
so be sure to do the diagnostic plots.

-npt_s- (or -vanelteren-) would still be appropriate, though, if your interest 
is in hypothesis testing.

Joseph Coveney

P.S.  About my previous post in this thread:  I've noticed that -npt_s- doesn't 
correct for ties when performing stratified analysis, so I don't recommend it 
now for the ordered-categorical response variable (at least until it gets 
fixed--I'll send an e-mail message to Peter Sasieni).  Use -vanelteren- 
instead.  The syntax for -vanelteren- is identical:  -vanelteren embryo_grade, 
by(culture_medium) strata(dam)-.  Either -npt_s- or -vanelteren- will be fine 
for the percentage response variable above, provided there are few, if any, 
ties.  If there are a substantial number of ties, then use -vanelteren- here, 
too, which corrects for ties.  Once -npt_s- is fixed to correct for ties in 
stratified analysis, then it will render -vanelteren- superfluous, because 
-npt_s- subsumes the latter's functionality.

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index