[William Lisowski]

I will try to make my post #13 clearer.

There is no way to combine two 10-digit numbers into a 20-digit number to serve as a dyadic-ID of the sort you claim to need - one that you can read and tell which two members are in the dyad - in Stata.

You will have to make do with the approach given in post #7 which yields an arbitrary number as a dyadic-ID. Note that your data will continue to include pidp and apidp so you will be able to identify the members of any given dyad, despite the arbitrary number assigned to it, for example

list pidp apidp if id==45378

[Giorgio Piccitto]

Dear William,
thanks for your answer, now it is clear.

I have only a doubt: by doing as you suggest, I will end up with a different ID along the years also for the same dyad (the relation between parent 345 and son 255 it will have dyadic-ID 1 in year A, dyadic-ID 2 in year B, for instance). This does not bias the analysis somehow? When I apply fixed effect model or other techniques of panel analysis, don't I run into some bias, given the fact that it is assigned a differend ID to the same dyad along time?

Thanks a lot, G.

[William Lisowski]

In post #13 above I addressed this. Append your seven waves of data into a single dataset, then with the appended dataset in memory

egen panel_ID = group(pidp apidp)

for all the data at once and the same panel_ID will be assigned to every observation with the same pair of pidp and apidp, regardless of year.

Do not create panel_ID seven times, once in each of the seven individual wave datasets. Then they will be different along the years, as you fear.

[Giorgio Piccitto]

Dear WIlliam,
thanks a lot, your tip leads me exactly to what I need, your help is very appreciated!

Best, g