It appears to me that:
- Every ID begins with a baseline record, each ID has only one baseline record, and each baseline record has (unsurprisingly) the baseline flag set to one. The programs below assumes so.
- But for DAILY records, baseline=1 ONLY if all the timepoint values match the preceding record. In your example the only daily records that fill this condition are the records immediate following the baseline record. Is that true for your data? Or can there be any other daily record that gets baseline=1 because its timepoint values match the preceding daily record? If so, this program will flag such baseline records.
The program below is untested (I'm not bothering to put your data into a DATA step):
data want;
set have;
by patientid;
array tp timepoint: ; /* Editted */
baseline=1; /* Default to 1 */
do over tp; /* If any timepoint difference, convert to 0 */
if lag(tp) ^= tp then baseline=0;
end;
if first.patientid=1 then baseline=1;
run;
If, however, the only daily records actually eligible for baseline=1 are the 2nd records of each patientid, whose timepoints match the baseline timepoints, then this more compact program will do:
data want;
set have ;
by patientid timepoint1 timepoint2 timepoint3 notsorted;
baseline=0; /* Editted additional statement */
if first.patientid=1 then baseline=1;
if lag(first.patientid)=1 and first.patientid=0 and first.timepoint3=0 then baseline=1;
run;
Notes:
- The "notsorted" keyword at the end of the BY statement tells sas that the data are grouped, but not necessarily in ascending order for all the by-variables.
- The "if first.patientid=1" test for whether the record-in-hand is the start of a patientid (which this program assumes is a baseline type of record).
- The "if lag(first.patientid)=1 and first.patientid=0" condition is true only for the 2nd record of each patientid (Note the "and first.patientid=0" part is required in case the preceding record was the only record for the preceding patientid).
and
the "and first.timepoint3=0" condition is true only when all the by-variables in the current record match their values in the following record.
