Do you need to recode and manipulate variables before or after multiple imputation?

[Jaycob Applegate]

I have heard conflicting advice on when to manipulate and recode variables. I was once told that you must impute on the most raw versions of variables and then recode (for example, turning a continous variable into categorical). I have taken this approach and am having convergence problems. However, my imputation and subsequent models work when I recode and then impute.

When I impure and then recode, my models are giving me error, saying there is variation between m = 1 and m = 2. Is there recommended guidelines here?

[Clyde Schechter]

I have heard conflicting advice on when to manipulate and recode variables.

That is because this is a controversial issue. There is no settled answer to this question, and prevailing opinion has flipped back and forth over time about this. If I recall correctly, the most recent fashion about this is to transform (recode) and then impute. But I might not be up to date on the latest oscillation as I don't use multiple imputation very often.

When I impure and then recode, my models are giving me error, saying there is variation between m = 1 and m = 2.

This probably means you are doing the imputation incorrectly. Variables that you will recode have to be -mi register-ed as -passive- before running -mi impute-. And then when you actually recode them you should use the -mi passive:- prefix and do the recoding with -generate-, -egen-, or -replace-, not -recode-. If you follow this protocol I think you will not encounter the error you are currently getting.