Incorrect results when subtracting time

[Dennis Ade]

Dear all

I am trying to compute call duration from subtracting two time periods in Stata but I get resultant incorrect time difference between the two variables. I have looked online but was unable to fix the issue. The "CallDuration" variable that I have has constant 12 hours added to it. I am trying to subtract 12 hours (the variable for this is "twlhour") from my "CallDuration" variable but the answer seems to be incorrect and unintuitive. Please tell me how to fix this problem.

This is my sample data:

* Example generated by -dataex-. To install: ssc install dataex
clear
input int(CalledNumber Date) double(Time CallDuration) float(twlhour call_dur)
15000 21304      -1893384240000      -1893455984000 4.32e+07 -1.893499e+12
15000 21304      -1893384240000      -1893455925000 4.32e+07 -1.893499e+12
15000 21304 -1893384300000.0002      -1893455891000 4.32e+07 -1.893499e+12
15000 21304 -1893384300000.0002 -1893455950999.9998 4.32e+07 -1.893499e+12
15000 21304      -1893384360000 -1893455985999.9998 4.32e+07 -1.893499e+12
15000 21304      -1893384660000      -1893455914000 4.32e+07 -1.893499e+12
15000 21304      -1893384660000 -1893455677999.9998 4.32e+07 -1.893499e+12
15000 21304      -1893384720000      -1893455813000 4.32e+07 -1.893499e+12
15000 21304      -1893384720000 -1893455954999.9998 4.32e+07 -1.893499e+12
15000 21304      -1893384840000 -1893455768000.0002 4.32e+07 -1.893499e+12
end
format %tdnn/dd/CCYY Date
format %tchh:MM_AM Time
format %tchh:MM:SS CallDuration
format %tcHH:MM:SS twlhour
format %tcMM:SS call_dur

I am running the following Code :

gen twlhour = 60*60*1000*12 // making a 12 hour variable
format twlhour %tcHH:MM:SS // formating
gen call_dur = CallDuration - twlhour //
format call_dur %tcMM:SS

Thanks anyways

[Nick Cox]

Always use double for any variable measured in milliseconds, So

gen double call_dur = CallDuration - msofhours(12)

There is no point to defining a variable containing a constant msofhours(12) or 60*60*1000*12. That value is repeated in every observation if you do that, It's not like a variable in a common or garden programming language, That's an entire variable (column) in your dataset.

See

help datetime

where the advice to use doubles is repeated often,

[Dennis Ade]

Thank you very much. This works! I will use doubles moving forward