|
|
Dear all,
I have a non-proportional regionally stratified random sample. Therefore, I have to weight the observations. Nevertheless, this causes a problem.
Some of my regions have a sample size of 90 or 100, but when I weight the sample, they account for 10 or 15 units.
I am trying to run a logistic regression where belonging to each region is estimated as a fixed effect.
My problem is that GENLIN command does not compute Betas for those regions with a weighted sampling size smaller than 100 individuals.
Of course, if I do not weight the sample I get estimators of betas for all regions. Nevertheless, weighting is important in order to get estimates for other covariates.
Does anyone know the syntaxis to avoid this problem and tell GENLIN to estimate a beta coefficient for all regions, even some of them have a small sample size.
Thanks for all
|
Locked
