generate duration variable

[Youngsang Lee]

Dear statalisters,

I've encountered a problem related to generating duration variable.
I already search answers of similar questions in statlist, but I can not solve this problem.
I want to make a 'peaceyear' variable like below Table.

dyad	year	dispute	peaceyear
2020	1950	0	1
2020	1951	0	2
2020	1952	0	3
2020	1953	1	4
2020	1954	0	0
2020	1955	0	1
2020	1956	0	2

Of course, I attemped the code in statalist.
The code is

gen peaceyear = .
bysort dyad(year): replace peaceyear = cond(dispute==0, cond(_n>1, peaceyear[_n-1]+1, 1),1)

However, I can not make the variable correctly.
This is my result.

dyad	year	dispute	peaceyear
2020	1950	0	1
2020	1951	0	2
2020	1952	0	3
2020	1954	1	1
2020	1955	0	2
2020	1956	0	3

Please notice what's wrong my state code.
If it is difficult, I will use excel and make peaceyear variable.
Please help me.
Thank you.

regards,
Youngsang Lee
Yonsei univ.

[Nick Cox]

I don't understand your rules. Your first year with no dispute is given peace year 1 and 0 in different observations and a year with a dispute is labelled peace year 4.

Compare these results.

* Example generated by -dataex-. For more info, type help dataex
clear
input int(dyad year) byte(dispute peaceyear)
2020 1950 0 1
2020 1951 0 2
2020 1952 0 3
2020 1953 1 4
2020 1954 0 0
2020 1955 0 1
2020 1956 0 2
end

gen wanted = dispute == 0
bysort dyad (year) : replace wanted = max(0, wanted[_n-1]) + 1 if wanted == 1 & wanted[_n-1] >= 1

list, sepby(dyad dispute)


     +-------------------------------------------+
     | dyad   year   dispute   peacey~r   wanted |
     |-------------------------------------------|
  1. | 2020   1950         0          1        1 |
  2. | 2020   1951         0          2        2 |
  3. | 2020   1952         0          3        3 |
     |-------------------------------------------|
  4. | 2020   1953         1          4        0 |
     |-------------------------------------------|
  5. | 2020   1954         0          0        1 |
  6. | 2020   1955         0          1        2 |
  7. | 2020   1956         0          2        3 |
     +-------------------------------------------+

Another approach is to use tsspell from SSC. tsspell is an otherwise unpredictable search term for many mentions of similar problems on Statalist. Here is how that works.

. tsset dyad year
. tsspell, cond(dispute == 0)

tsspell creates several variables, of which _seq is identical to wanted above.

See also https://www.stata-journal.com/articl...article=dm0029 for a more general discussion.