03-19-2018 04:44 PM
I am looking to lag the exposure to outcome by a year. What I mean by this is that for a given year in my data, I want the exposure in one year to be associated with the outcome in the next, instead of the same year. This is because I want to avoid reverse causation.
So if the exposure for 1999 is 26, I would want it to match with the outcome for 2000.
Here is my code without the lag:
PROC GENMOD data=data1; CLASS StateID year (ref=first) event (ref=first); MODEL Outcome =Exposure confounder / DIST=poisson LINK=log TYPE3 WALD offset=lnpop; REPEATED SUBJECT=stateid; run;
I have an offset of state population and a repeated subject on stateID. Outcome is my main outcome and exposure is my main exposure. Confounder is a confounder. I do not think that the confounder should be lagged, just the main exposure and outcome.
Thanks so much!
03-19-2018 05:46 PM
You could run your analysis on table data2 given by:
proc sql; create table data2 as select a.*, b.Exposure as lastYearExposure from data1 as a inner join data1 as b on a.stateId = b.stateId and b.year = a.year-1; quit;