Dear Amelia Developers
I have the following situation. Lets take two independently collected
samples from the same population (drawn with the same sampling
technique, with low chance of overlap and data collected around the
same time). Sample 1 has variables A and C-E. Sample 2 has
variables B-E. Both samples have 500 observations. I merge the
datasets so I get a dataset with 1000 observations with variables A
through E (where A-E are continuous normal). A and C-E is observed
for the first 500 cases and B-E for case 501-1000. I can assume that
data is missing completely at random. My imputation model would
include all variables (from A-E) and in my analysis model I want to
regress A on B for example (correlate A with B). I just talked to
Craig Enders and he verified that this imputation will not work with
Proc MI or NORM as the maximum likelihood model is not identified due
to no cases where A and B are observed at the same time. But I know
Amelia uses a different procedure. Is it possible to run this in
Amelia and get unbiased results?
Thanks
Levi
-
Amelia mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive:
http://lists.gking.harvard.edu/?info=amelia
More info about Amelia:
http://gking.harvard.edu/amelia