Hi!
I am using the new MI software in Stata 11 and I am running into some
problems with the estimation commands. I was wondering if anybody had
any ideas on how to get around the specific problem of imputing a
variable with a lot of data missing which you will later use as a
grouping variable in your estimation models.
let me explain this better. For my analysis I fit a ton of models by
income quintiles. However, I have a lot of data missing for income,
and so I imputted the dataset using MI creating 40 complete datasets.
For each of the datasets MI has substituted the missing value for
income with a possible value. The problem is that when I fit models
for different quintiles, Mi estimate clearly recognizes that the
estimation sample varies from one imputed dataset to another, as
follows:
. mi estimate : regress outcome predictor predictor2 if income<= XXX
estimation sample varies between m=1 and m=2
To make the model work I asked MI to ignore the variation across
estimation samples by using esampvaryok. However, to make matters
worse, I then need to add fixed effects to deal with the clustering of
my data, and then MI can't fit the model at all because different
samples include different fixed effects.
. mi estimate : regress outcome predictor predictor2 FE1-FE100 if income<= XXX
mi estimate: omitted terms vary
The set of omitted variables or categories is not consistent
between m=1 and m=2; this is not allowed. To identify varying sets,
you can use mi xeq to run the command on individual imputations
or you can reissue the command with mi estimate, noisily
Any ideas on how to deal with this problem?
Clara
*
* 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/
