Filling in all possibilities in longitudinal dataset

[Chris James]

I have data in the following form

clear
input personid date charityid gave charity_open_date
1 18265 1 1 18263
1 18267 2 1 18263
1 18273 3 1 18272
2 18263 1 1 18263
2 18273 2 1 18263
3 18264 1 1 18263
3 18271 2 1 18263
end

I would like to create every possible personid-charityid pair in the data. The logic I envisioned would establish a start date, the minium date for each personid, and an end date, the maximum date for each personid. Then, if a charityid is open in this window (if start_date <= charity_open_date & end_date >= charity_open_date) and they did not give to this charityid (gave == 1), an entry would be created filling in the personid, charityid, and gave == 0. All other variables would be missing (could be string or float missing).

In this example, personid == 1 donated to all charities so no change.

personid == 2 could have donated to charityid == 3 and did not so it would add one observation

2 . 3 0 .

personid == 3 leaves the data before charityid 3 opens so there is no change to their records.

[Clyde Schechter]

-help fillin-

[Marcos Almeida]

I don't know the logic of doing this, for Stata can cope with the analysis without having to "filling the gaps". That said, I wonder whether - tsfill - wouldn't provide the solution you wish.

[Nick Cox]

As a footnote to #2 note that https://www.stata-journal.com/sjpdf....iclenum=dm0011 on fillin has giving to charity as an explicit example.