Re: st: mancova with weighted survey data?

[Steve Samuels <sjsamuels@gmail.com>]

I realize this is an old post.

Billy's suggestion of -mvreg- would be my first choice, since MANCOVA is
a special case of multivariate regression, just as ANOVA is a special
case of single outcome regression. However, -mvreg- does not take
probability weights and is not survey-aware, so cannot compute proper
standard errors and tests.

However Dennis Roodman's package -cmp- (from SSC) can do the same thing
as -mvreg-, does take probability weights, and *is* -svy- aware. Look
way down in the -help- for an example that shows equivalent -mvreg- and
-cmp- analyses. I haven't used -cmp- so I can't advise you on specifics,
but it should solve your problem.



Steve
sjsamuels@gmail.com
> 
> On Sep 5, 2013 at 8:48 AM William Buchanan wrote:
>  
> Hi Annie,
> 
> Why not use multivariate regression instead?  Another option, that could potentially be a bit more beneficial is to consider using a multiple group MIMIC/CFA and test the different in the latent means of "stigma" between your two groups.  If you've searched around the net much you'll probably come across several different threads regarding ANOVA with probability weights or other survey weights in Stata and other packages.  
> 
> HTH,
> Billy
> 
> 
> 
> On Sep 5, 2013, at 9:36 AM, Fox-Galalis, Annie B. wrote:
> 
> Hello Statalist!
> 
> I am very new to Stata and am using version 10.  I need to run a MANCOVA  with 5 DVs (5 different measures of stigma), one 2-level IV (PTSD or No PTSD), and a continuous covariate (social desirability), but using weighted survey data.   We are trying to figure out if people with PTSD have higher levels of stigma than people without PTSD, controlling for social desirability.  Correlations between the stigma measures vary between .37-.63.
> 
> For other analyses, I have used the -svy- command, but have quickly figured out that I cannot do the -manova-  with -svy-.  I have done some digging in various help menus and google, but haven't yet found a workaround for how to do this analysis.   I thought that I might be able to do the regular -manova- command and add the 
> -pweights- command,  but that didn't work/isn't allowed by the program.  
> 
> If anyone has any suggestions, workarounds or syntax for how I could do this analysis, it would be greatly appreciated!
> 
> Thanks!
> 
> Annie Fox 
> 
> 
> 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/