Keeping the first date/observation every 24 weeks

[Pawel Czarnowski]

Hello everyone,

I have a dataset of thousands of companies and the dates of hundreds of announcements made by each company. Variable "id" identifies a company. Variable "date" represents the announcement date.

Please see my data below:

* Example generated by -dataex-. To install: ssc install dataex
clear
input long id int date
1 7550
1 7578
1 7704
1 7900
1 7921
1 8145
1 8320
1 8600
1 8663
1 8677
1 8726
1 8901
1 8985
1 9104
1 9209
2 7396
2 7403
2 7592
2 7732
2 7851
2 7928
2 8040
2 8187
2 8194
2 8299
2 8362
2 8565
2 8628
2 8726
2 8740
end
format %d date

I want to keep/identify the first announcement every 24 weeks, for every company.
For example, in my data above, for the first id, I am interested in keeping/identifying 02sep1980 (7550), 18aug1981(7900), 20apr1982(8145), 12oct1982(8320), 19jul1983(8600), 15may1984(8901), 04dec1984(9104).

In other words, I want to keep subsequent announcements that are as close to each other as possible, but not any closer than 24 weeks apart.

I would be very grateful for your help. Please let me know if I should clarify my problem any further.

Pawel

(I use Stata/SE 15.1)

[Clyde Schechter]

local 24_weeks = 7*24

by id (date), sort: gen ref_date = date + `24_weeks' if _n == 1
by id (date): replace ref_date = cond(date > ref_date[_n-1], date + `24_weeks', ref_date[_n-1]) if _n > 1
by id (date): keep if ref_date != ref_date[_n-1]

Thank you for using -datatex- on your very first post!

[Pawel Czarnowski]

Thank you very much, that worked perfectly.

[Nick Cox]

panelthin from SSC is I think relevant here. It's also a search term for mentions of this problem on Statalist.