st: Warning: variance matrix is nonsymmetric or highly singular

["Butler, Lauren" <lauren_butler@unc.edu>]

Hi, 

When I run the following code:
 
svy, subpop(subpop if adult==1): logistic drinker ib(1).gender##i.wave ib(5).agegrp ib(1).race ib(4).edugrp ib(3).poverty ib(1).bmicat ib(1).weekend 

margins i.gender#i.wave,  subpop(subpop if adult==1) vce(unconditional) atmeans post pwcompare(group) mcompare(bonferroni)

I get this error after the margins command: “Warning: variance matrix is nonsymmetric or highly singular”

and then I get estimates but no standard errors. 

Note that this error does not occur after running the logistic model and this error does not occur if I remove the "vce(unconditional)" from the margins command statement.  

Additionally, rarely, the margins will run without this getting this warning message.  Does anyone have experience with this warning and how to address it?

I've already taken a look at previous STATALIST threads on this topic and they did not help.

Thanks,


Lauren Butler, RDN, LDN
Doctoral Student, Nutrition Epidemiology
Department of Nutrition
UNC Gillings School of Global Public Health Carolina Population Center
123 W. Franklin St University Square
CB#8120, Chapel Hill, NC 27516
Phone: 919 951 4345
Email: lauren_butler@unc.edu
*
*   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/