non-repetitive ratio?

[River Huang]

Dear All, I found this question here (in Chinese). The data set is:

* Example generated by -dataex-. For more info, type help dataex
clear
input str4 citing float year str4 cited
"A" 2018 "C"
"A" 2018 "D"
"A" 2019 "C"
"A" 2019 "F"
"A" 2020 "E"
"A" 2021 "B"
"A" 2021 "C"
"A" 2021 "D"
end

For each "citing" company (and assume there are many companies) and each "year", we'd like to obtain a non-repetitive ratio, which is explained below.
For citing=A, and year=2021, there are three "cited" companies (B,C and D). One can find that only B does not appear in the previous 3 years. As a result, the non-repetitive ratio is 1/3 (for citing=A, and year=2021). Any suggestions? Thanks.

[Øyvind Snilsberg]

maybe,

bysort citing cited (year): gen x = cited==cited[_n-1] & (year-year[_n-1])<=3
bysort citing year: egen wanted = mean(x)
replace wanted = 1 - wanted if wanted

[River Huang]

Dear Øyvind, Thanks a lot for this helpful suggestion.