Hi, I was hoping that someone at Statalist could help me with this. I’m trying to use multilevel logistic regression with complex survey data. I’m using the Health and Retirement (HRS) study to examine longitudinal data but I’m unsure whether I’m specifying the complex survey design correctly. Below is a description of the variables and how I’m using the svyset and melogit commands.
Stratum: raestrat
PSU: raehsamp (which indicates whether or not it is a face to face interview or telephone interview, waves alternate between telephone and face to face interviews)
Individual unique identifier: hhidpn
Sample weight for individual: wt
According to the HRS documentation, analysis of cross-sectional data should be specified as follows
svyset raehsamp [pweight=wt], strata (raestrat)
Since I’m examining longitudinal data I’m specifying the model this way
svyset raehsamp, strata(raestrat)
svy: melogit hypertension c.age##i.race || hhidpn:, pweight(wt)
The above specification seems to work but I’m unsure if it is correct
Alternatively, I’ve read that the model should be specified as follows, but I’ve never been able to get this to work
svyset raehsamp, strata(raestrat) || hhidpn, weight(wt)
svy: melogit hypertension c.age##i.race || raehsamp: || hhidpn:
Any help with this would be appreciated.
Thank you,
Brandon
0 Response to Multilevel models with complex survey design
Post a Comment