Propensity Matching--Selecting Treatment=1 vs Treatment=0

krm14 — Mon, 05 Oct 2020 19:55:43 GMT

Hello,

I have two groups defined by whether or not they received an intervention. I want to propensity match the groups on age, sex, race, risk score and facility size. Group 1 is the treated group (n=274,678), and Group 2 is not treated (n=131,799).

I tried the following code first:

ods graphics on;
proc psmatch data=inp_tbi_PS region=cs;
class intervention BENE_RACE_CD GNDR_CD fac_size;
psmodel intervention(Treated='1')= AGE GNDR_CD BENE_RACE_CD fac_size risk_score;
match method=optimal (k=1) stat=lps caliper=0.25;
assess lps var=(GNDR_CD fac_size) / weight=none plots(nodetails);
output out (obs=match)=inp_tbi_PS_out lps=_Lps matchid=_MatchID;
run;

I got an error message saying that "A feasible optimal fixed ratio matching that has the specified parameters does not exist".

Since the treated group is larger than the non-treated group, should I change my code to "Treated=0"?

Thanks in advance!

Re: Propensity Matching--Selecting Treatment=1 vs Treatment=0

RichardDeVen — Mon, 05 Oct 2020 23:04:17 GMT

Are the age values actual age or an age category value such as 1=<18, 2=18 to 29, 3=30-39, etc

Re: Propensity Matching--Selecting Treatment=1 vs Treatment=0

krm14 — Tue, 06 Oct 2020 01:44:51 GMT

@RichardDeVen Actual age

topic Re: Propensity Matching--Selecting Treatment=1 vs Treatment=0 in Statistical Procedures

Propensity Matching--Selecting Treatment=1 vs Treatment=0

Re: Propensity Matching--Selecting Treatment=1 vs Treatment=0

Re: Propensity Matching--Selecting Treatment=1 vs Treatment=0