Creating lagged variables for panel data with multiple imputations

[alex badalyan]

Hi, I have panel data set by countrycode and year, with multiple imputations.

When I try to sort data and command:

sort countrycode year

for some reason the imputed dataset loses variability, i.e. all the observations for non-imputed variable are the same. e.g. for US in years 2000-2010 I imputed var1. When I sort by countrycode and year, var2, var3 etc are just the same value for all US observations in 2000-2010.

Now to what I was actually doing this for: trying to create lags of my variables.
What I had in mind for code was:

sort countrycode year
by countrycode: gen lagvar1 = var1[_n-1] if year==year[_n-1]+1

As mentioned above, that problem with sorting arises. Is it possible to create lagged variables for imputed data and if so, what is the code?

[William Lisowski]

Way over my head here but maybe this will start some discussion.

If I were fitting panel data I'd be using xtreg and handling lagging using time series variable lists (help tsvarlist). Since help xtreg supports mi estimate that would seem to be a possible direction.