st: Bootstrap results of margins in multiply imputed data

["Singer, Adam" <asinger@prgs.edu>]

Dear Statalist,

I am bootstrapping the percent change in x between two years using Stata 12.1.  I have written the following program and successfully called it in non-imputed data with the last line of code:


program pctchange, rclass
     logit x i.year
     margins year
     mat A = r(b)
     return scalar ratio = A[1,2] / A[1,1]
end

bootstrap r(ratio), reps(200) seed(12345): pctchange


My question pertains to implementing this program in multiply imputed data.  I have a multiply imputed dataset in Stata's mi format.  In this dataset, the regression must be run using "mi estimate: logit x i.year".  However, -margins- does not recognize regression results obtained using -mi estimate-.  In order to account for that and other research I've done on user-defined programs in multiply imputed data, I have rewritten the program as follows:


program pctchange, eclass properties(mi)
     logit sigpain i.year
     margins year, post
     mat A = e(b)
     ereturn scalar ratio = A[1,2] / A[1,1]
end

bootstrap e(ratio), reps(200) seed(12345): mi estimate, cmdok: pctchange 1


The last command doesn't work.  How do I combine -bootstrap- and -mi estimate-?  Is there a better way to write the program to accomplish my goal?  In addition, if I remove -bootstrap- from the command and simply use "mi estimate, cmdok: pctchange 1" (which works), I notice that e(ratio) is empty (when I ran my first program in non-imputed data, r(ratio) was not empty).  Does -mi estimate- interfere with my program's ability to return scalars?

Thank you very much!

Adam Singer

__________________________________________________________________________

This email message is for the sole use of the intended recipient(s) and
may contain confidential information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/