st: Imputing categorical and continuous variables simultaneously

[Mosi Ifatunji <ifatunji@gmail.com>]

All,

I am currently working on multiple imputation. I have completed all of the programing to generate 20 imputed datasets and was going to begin post hoc coding when I thought to send this question.

The variables that I am imputing (in both the substantive and imputation models) are both categorical (i.e., Likert scales) and continuous variables (e.g., income and years of education). According to my read of the 'mi' command set in Stata 11, one can use the 'mi impute ologit' for the former and 'mi impute mvn' or 'mi impute regress' for the latter. However, if you are imputing variables that are both categorical and continuous, the only imputation method is 'mi impute mvn.' Is this an accurate reading of the materials? Or, is it possible to specify different imputation functions (e.g., ologit and mvn) for the same session?

The implication is that, when I use only 'mi impute mvn' I am forced to recode categorical variables post hoc. If I were able to impute these variables using 'mi impute ologit' while simultaneously imputing variables that are continuous with 'mi impute mvn', it would be a big time saver. If there is a way to simultaneously impute both types of variables, how might I do so and are there any good readings on the matter (say in Stata Journal) that you might recommend?

All my very best,

Mosi

-- 
Mosi A. Ifatunji, Ph.D. Candidate
Race and Difference Initiative
Dissertation Research Fellow
Emory University
201 Dowman Drive
Atlanta, Georgia 30322
Cell/text: (312) 607-2825
Email: mosi@ifatunji.com
Web: http://www.ifatunji.com/
Twitter: @ifatunji

"Power concedes nothing without a demand."
                          - Frederick Douglass, 1857


*
*   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/