st: Differences in -mi ice- command between Stata 12.1 and 13

[Mosi Ifatunji <ifatunji@gmail.com>]

Colleagues,

I upgraded from Stata 12.1 IC to Stata 13 SE yesterday afternoon. This morning I re-ran an imputation command without changing anything in the syntax. I used the -mi ice- command because I wanted to use features of the older ice command while producing datasets that were usable with the newer -mi- command set (see http://goo.gl/MmJHRr). I was particularly interested in producing estimates using the link-ed chain approach.

My query is that when running the command in Stata 12.1 IC everything ran smoothly. Running the exact same syntax in Stata 13 SE resulted in -auglogit- regression for one of the variables in three of the twenty datasets generated. I am interested in any explanations for this inconsistency.

I should say that this is not my first time getting an -auglogit- regression during multiple imputation. In the past though, if I get -auglogit- for a variable, I got it for that same variable in the production of all datasets. I then worked to respecify my model so as to not get any -auglogit- defaults. This case is different because -auglogit- is only being used to produce certain datasets (3/20; for the same variable) and the imputation model that I used before, and that did not produce -auglogit-, is now producing -auglogit- ; same variables same syntax.

Any ideas?

-- Mosi



*
*   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/