Date:         Fri, 18 Dec 2009 12:38:44 -0800
Reply-To:     John <epijohn@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         John <epijohn@HOTMAIL.COM>
Organization: http://groups.google.com
Subject:      Proc QLIM and endogeneity with survey data
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

Hi,

I'm struggling with analyzing two large surveys on malaria parasitemia
and how the interventions (spraying and net coverage) the effect
parasitemia prevalence (binary outcome). These are 2 independent
sample 2-stage cluster surveys with strata, similar to DHS.
In proc surveylogistic, I have developed a model which shows reduced
parasitemia by increasing level of the intervention, but when I look
for endogeneity amongst the levels of intervention, my rho (using proc
qlim with binary probit for each level one at a time with outcome) is
significant.

How do I handle this? Seems like I still need to work this model for
variance estimates (of the survey design) through surveylogistic. What
to do next?

Thanks.