st: RE: generalized Dunnett's test?

["Airey, David C" <david.airey@vanderbilt.edu>]

.

I was hoping for a closed solution, as this is already in the context of a lengthy simulation. Thanks for the reference! I think I will take the easy route and compare Holm's and Dunnett's approaches first to see if there is a big difference.


> It would involve some programming work, but you could try resampling methods, as described in Westfall and Young (Resampling-based Multiple Testing). 
> 
> 
> Al
> 
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [
> mailto:owner-statalist@hsphsun2.harvard.edu
> ] On Behalf Of Airey, David C
> Sent: Tuesday, January 18, 2011 8:17 AM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: generalized Dunnett's test?
> 
> .
> 
>> Dunnett's test is an ANOVA post-hoc test, a many-to-one pairwise test, typically used for comparing many experimental treatments each to a single control following a oneway ANOVA. Is there a generalization of this test for likelihood ratio tests? I have a full model and a set of reduced models (NOT from oneway ANOVAs), where each reduced model constrains the parameters of the control and one treatment group to be shared/equal. Each LRT for the full and a reduced model tests a control vs treatment difference. I can apply a multiple test adjustment to the vector of LRT p-values, like Holm's adjustment, but I'm curious if there is a Dunnett analogue for this situation. Some multiple test procedures assume independence of the tests, but there is the same control group in each pairwise test in Dunnett's situation, and in my set of LRTs. Although I've not simulated it, I'm assuming with many tests, Dunnett's procedure will beat Holm's correction for a many-to-one set of pairwis!
 e c!
>>  omparisons.

*
*   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/