I used the -ice- routine for multiple imputation on my data. For some
of my models, I want to run separate models for those students who
applied to college and those who did not apply to college. When I run
the models using an "if apply==0/ if apply==1" and -micombine-, my
output tells me that there are different sample sizes in imputation 2
through 5 compared to imputation 1. I have believe these sample sizes
are different because of different imputation values within one case
(i.e. one person has a 0 1 0 0 1 for their five values for apply,
which then changes the sample sizes depending on the imputation number
when I restrict the models to apply==1). What is the technically
correct way to deal with this? Do I go ahead and use the regressions
that have different sample sizes across imputations or do I restrict
my sample using a non-imputed measure of apply or is their another
method?
Thank you for your help,
Elizabeth Covay
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
