Creating new variable which sums up the values of another variable

[Alex Caridyas]

Hi,

I would like to calculate the total number of months spent in prison by summing up the time in months spent for each unique “episode” in prison. I have three variables

• Unique person ID
• Prison episode
• The number of months spent in prison

The problem is that I do not want to recount the months. For e.g. individual B had two episodes in prison, in the first episode they spent 10 months and in the second episode they spent a total of 16 months (4 months when they were observed the second time and then 16 months by the time, they were observed a third time). I am not sure how to tell Stata to take the maximum “months in prison” value for each unique episode in prison and add them up.

I have tried the following:

bysort ID prison_ep: egen max=max(months)

which creates a maximum column, but I am not sure how to then add up the total month without double counting.

The column I am trying to produce is total_months (I have manually calculated it in the data example below to demonstrate)

* Example generated by -dataex-. To install: ssc install dataex
clear
input str1 ID byte(prison_ep Months) float(max total_months)
"a" 1  2 26 26
"a" 1 14 26 26
"a" 1 26 26 26
"b" 1 10 10 26
"b" 2  4 16 26
"b" 2 16 16 26
"c" 1  6  6  6
"d" 1 10 10 25
"d" 2 10 10 25
"d" 3  2  5 25
"d" 3  5  5 25
end

[George Ford]

bys ID: egen sum_months = sum(cond(Months==max),Months,.)

[Alex Caridyas]

Thank you, George.

[Nick Cox]

George Ford gave good advice; note, however, that as from Stata 9 the equivalent function is documented as total(). But sum() continues to work.