Hi,
The context of my question is that I am trying to estimate average, median or mean, time tourists spend abroad and average number of times they spend by analyzing their credit card history abroad.
My data is in the following format where one row represents one vacation period for one cardholder.
- Card_ID
- Last_Foreign_Trans_Dt
- Last_Foreign_Trans_Dt
- Left_Truncated ={0 if there exists a domestic transaction in the observation period prior to the vacation, 1 if a domestic transaction in the observation period does not exist prior to the vacation}
- Right_Censored ={0 if there exists a domestic transaction in the observation period after the vacation, 1 if a domestic transaction in the observation period does not exist after the vacation (i.e. cardholder is assumed to be still on vacation)}
- Time ={Last_Foreign_Trans_Dt - Last_Foreign_Trans_Dt, in days}
- Transactions ={Number of transactions made in vacation period}
As a consequence of how the data is formatted, Card_ID can be repeated (the same Card_ID on a different row represents a different vacation period for that cardholder).
My question is how do I estimate time spent abroad given that some vacations are left-truncated (i.e. unsure when they started) and some are right-censored (they are still on vacation). If the data was only right-censored and did not feature any left-truncation then PROC LIFETEST would be suitable. I've read, however, that PROC LIFETEST cannot deal with left-truncation but that PROC PHREG can.
Does anyone know please how I should approach this, should I be using PROC PHREG?
