Say I have two datasets A & B. They have the same variables in each except that set A has one additional variable.

Would it be valid to impute the variable for dataset B, and then to only conduct analysis on dataset B?

For example:

Code:
webuse lbw, clear
set seed 12345
generate wave = round(uniform(),1)
replace bwt = . if wave==1 //artificially creating missing data in half of dataset

mi set mlong
mi register imputed bwt
mi impute chained (pmm, knn(5)) bwt = age lwt i.race, add(5) rseed(12345)
mi estimate : glm age lwt i.race bwt if wave==1 //command runs, but is only conducted on part of sample where all data was missing originally