topic Re: Proc PS Match not using all the input data in Statistical Procedures

Proc PS Match not using all the input data

riyaaora275 — Wed, 05 Aug 2020 14:06:16 GMT

I have used the following code -

proc psmatch data = x;
class a b c cate;
psmodel cate(Treated = 'On') = a b c d ;
match method = greedy(k=1)  ;
assess lps var=(a b c d )/plots = all weight = none;
output out(obs=match)= y;
run;

however the treated group has 6000 obs and control has 8000 obs , then why is the number of obs used 1134 and 1411 ?

Re: Proc PS Match not using all the input data

Reeza — Wed, 05 Aug 2020 15:04:54 GMT

Do you have missing values in your variables, specifically a, b,c, d?

@riyaaora275 wrote:

I have used the following code -
proc psmatch data = x;
class a b c cate;
psmodel cate(Treated = 'On') = a b c d ;
match method = greedy(k=1)  ;
assess lps var=(a b c d )/plots = all weight = none;
output out(obs=match)= y;
run;
however the treated group has 6000 obs and control has 8000 obs , then why is the number of obs used 1134 and 1411 ?

Re: Proc PS Match not using all the input data

riyaaora275 — Wed, 05 Aug 2020 16:14:26 GMT

yes

Re: Proc PS Match not using all the input data

Reeza — Wed, 05 Aug 2020 17:40:07 GMT

Any row with missing data for a, b, c, or d will be excluded from the analysis. If you need it included, there may be an option to account for that or you can recode them to a different value if that makes sense given the context.

Or you can impute them ahead of time. There are multiple options but it definitely requires subject matter expertise.