Creating a variable that gives the day from baseline based on a date variable

[Patrick Kafeero]

Hello,
Im kindly requesting on guidance on the issue below.
I have data in long format ie IDs of patients with their dates of LP. Each observation indicates a different date of LP and the variable DateLp shows the difference.
I want to be able to generate a variable DayOfLP which basically counts the number of days from the first date of LP. e.g if DateLP=12SEP2021 and DayOfLP is 1, then DateLP of 14SEP2021 would result to DayOfLP being 3.
An example dataset is below.

* Example generated by -dataex-. For more info, type help dataex
clear
input str6 PatId int DateLP
"64-001" 19475
"64-001" 19478
"64-002" 19478
"64-004" 19492
"64-004" 19495
"64-004" 19499
"64-004" 19507
"64-006" 19505
"64-006" 19507
"64-006" 19512
end
format %td DateLP

I have tried the code

quiet by PatId: gen DayOfLP=(_N-_n) if DateLP!=.

but this generates numbers that are like tags for the duplicates (starting from 0 onwards)

I'd be grateful if some guidance is given.

Thanks.

[Rich Goldstein]

If I understand you correctly, and if no ID appears more than twice, then I think this will give what you want:

quiet bys PatId (DateLP): gen DayOfLP=DateLP - DateLP[_n-1] if PatId==PatId[_n-1]

[Clyde Schechter]

by PatId (DateLP), sort: gen DayOfLP = DateLP - DateLP[1] + 1

Added: Crossed with #2. This code will produce correct results no matter how many times the same PatID appears.

[Rich Goldstein]

Note that Clyde Schechter and I read your original post somewhat differently as he adds "1" to each difference and I don't - presumably you can choose which, if either, is actually what you want

[Patrick Kafeero]

by PatId (DateLP), sort: gen DayOfLP = DateLP - DateLP[1] + 1

Added: Crossed with #2. This code will produce correct results no matter how many times the same PatID appears.

Thanks a lot Clyde.
This was very helpful.

[Patrick Kafeero]

Note that Clyde Schechter and I read your original post somewhat differently as he adds "1" to each difference and I don't - presumably you can choose which, if either, is actually what you want

Thank you Rich. I think I can now understand the difference and its application as I have tried your code and Clyde's.

Thanks.