topic PROC PSMATCH - Optimal matching in Statistical Procedures

PROC PSMATCH - Optimal matching

KK_Mangla — Mon, 29 Jun 2020 04:20:49 GMT

Error: A feasible optimal fixed ratio matching that has the specified parameters does not exist.

I am trying to do propensity score matching using method = optimal and caliper=0.25, however, I keep getting an error saying "A feasible optimal fixed ratio matching that has the specified parameters does not exist". When I change use caliper=., it works fine.

I had an understanding that if caliper value is unspecified, the default caliper of 0.25 is used. However, providing value of 0.25 in caliper explicitly gives me an error. Can I still say that Caliper=. means that the caliper value considered while matching is the default value of 0.25?

Re: PROC PSMATCH - Optimal matching

ballardw — Mon, 29 Jun 2020 05:20:41 GMT

1) show the entire code you are using. Since we don't know any of the other options you might be using it would involve lots of guessing.

2) and may be related to contents of your data, especially if you specify lots of constraints compared to the number of records available.

Re: PROC PSMATCH - Optimal matching

KK_Mangla — Mon, 29 Jun 2020 05:31:43 GMT

Below is the code I am using. Note that the prop_score is calculated separately using PROC LOGISTIC and i have tried both OPTIMAL and VARRATIO method

Title 'PSM using score difference';
proc psmatch data=work.&data_set. region=cs;
class &cat_var.;
psdata treatvar = case_cntrl_fl(Treated='1') PS=prop_score;
match method=varratio(kmax=1) exact=(&exact_var.) stat=ps caliper(mult=stddev)=0.25;
assess lps ps var=(prop_score sex &var_list.) / plots = (boxplot);
output out(obs=match)=WORK.psm ps=ps lps=lps matchid=_MatchID matchwgt = _MATCHWGT_;
run;

Re: PROC PSMATCH - Optimal matching

MichaelL_SAS — Mon, 29 Jun 2020 13:59:05 GMT

As described in the CALIPER= syntax section of the PROC PSMATCH documentation, specifying "CALIPER=." requests no caliper constraint be applied to the matching problem. Specifying CALIPER=. is different then not specifying the CALIPER= option, in which case you are correct CALIPER=0.25 would be used as the default.

Re: PROC PSMATCH - Optimal matching

KK_Mangla — Mon, 29 Jun 2020 13:58:24 GMT

Thank you so much!

Re: PROC PSMATCH - Optimal matching

MichaelL_SAS — Mon, 29 Jun 2020 13:59:26 GMT

Glad to help.