Good morning
I am trying to solve an issue with a big dataset (unbalanced panel with several missing data).
The task is quite easy, but the dimension of the panel require an automated routine.
Briefly speaking, I need to calculate a moving average of a variable including 3 lags and the current value.
I find that the command

tssmooth ma movaverage = targetvariable, window(3 1)

works fine, since it manages very well the missing values but it has a drawback; it provides values for each year, including the first three ones.
However, I expect (need) three missing values at the beginning of each time series, since only in year 4 the average can be properly calculated.
Some Statalist posts suggest to use the

(L3.var + L2.var + L1.var + var)/4

strategy, but I guess that missing data create several problems here.
Is there a better strategy to manage the issue?
Thanks again as usual for your insightful comments
Simone