topic Re: PSMATCH Procedure in SAS Programming

PSMATCH Procedure

sasphd — Mon, 20 Dec 2021 14:56:47 GMT

Hello,

I want to condunct PSMATCH to have the matched sets of id which correspond to the treatment group.

I want your help to understand Example 96.4 Greedy Nearest Neighbor Matching https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.3/statug/statug_psmatch_examples04.htm

column match_id is supposed to give the id of matched group so why I did not see sudentID in this column???

Re: PSMATCH Procedure

Tom — Mon, 20 Dec 2021 15:31:48 GMT

Doesn't that report mean that StudentIDs 156 and 173 form the first match?

Re: PSMATCH Procedure

sasphd — Mon, 20 Dec 2021 15:39:17 GMT

so why I did not find the same thing using my data and program ?

proc psmatch data=man.pms_data region=treated;

class IPO ;

psmodel IPO(Treated='yes')= age Net_asset_share_class Expenses Turnover manger_tenure_average count_name flux;

match distance=lps method=greedy(k=1) caliper=0.5;

assess lps var=( age Net_asset_share_class Expenses Turnover manger_tenure_average count_name flux)

/ stddev=pooled(allobs=no) stdbinvar=no

weight=none;

output out(obs=match)=OutEx4 matchid=_MatchID;

run;

this is my OUTEX4

Re: PSMATCH Procedure

maguiremq — Mon, 20 Dec 2021 15:40:32 GMT

I've never used the PSMATCH procedure, but @Tom seems to be right. I looked up the user guide and it seemed to go along with that idea.

https://support.sas.com/documentation/onlinedoc/stat/142/psmatch.pdf

Page 7687

and the MATCHID=_MatchID
option creates a variable named _MatchID that identifies the matched sets of observations.

Re: PSMATCH Procedure

maguiremq — Mon, 20 Dec 2021 15:45:22 GMT

Are those all the variables in your data set? It's hard for us to tell from a screenshot. Maybe run a PROC CONTENTS just to show the variables so we can be certain it's not actually in your data set.

Re: PSMATCH Procedure

sasphd — Mon, 20 Dec 2021 15:59:13 GMT

yes this is all my variable by not all the data. this is a screenshot of proc contents

Re: PSMATCH Procedure

maguiremq — Mon, 20 Dec 2021 16:01:43 GMT

Thanks, but this appears to be your original data set. What about `OutEx4`? Can you show me that one?

With this statement:

output out(obs=match)=OutEx4 matchid=_MatchID;

you are creating a new data set, `OutEx4`. That should have _matchID.

Re: PSMATCH Procedure

sasphd — Mon, 20 Dec 2021 16:25:53 GMT

thanks for your help

the outExt give the matched Id. when I make a proc sort by _MatchID I find the same table as Output 96.4.14: Output Data Set with Matching Numbers

however I find that the program match each observation for the treatment group with only one observation in the control group

but I want to match each observation for the treatment group with multiple observations in the control group if the PMS score allow.

can you help me to have the good procedure for that ?

Re: PSMATCH Procedure

Reeza — Mon, 20 Dec 2021 16:51:49 GMT

@sasphd wrote:

thanks for your help

the outExt give the matched Id. when I make a proc sort by _MatchID I find the same table as Output 96.4.14: Output Data Set with Matching Numbers

however I find that the program match each observation for the treatment group with only one observation in the control group

but I want to match each observation for the treatment group with multiple observations in the control group if the PMS score allow.

can you help me to have the good procedure for that ?

You specified K=1 in your code, which means it matches 1:1.

To match at a different ratio you would specify a different k.

 

proc psmatch data=man.pms_data region=treated;

   class IPO ;

   psmodel IPO(Treated='yes')= age Net_asset_share_class Expenses Turnover manger_tenure_average count_name flux;

   match distance=lps method=greedy(k=1)  caliper=0.5;

   assess lps var=( age Net_asset_share_class Expenses Turnover manger_tenure_average count_name flux)

          / stddev=pooled(allobs=no) stdbinvar=no

             weight=none;

   output out(obs=match)=OutEx4 matchid=_MatchID;

run;

https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_psmatch_syntax08.htm#statug.psmatch.methodopt

METHOD=GREEDY <(K=k ORDER=order-option)>

requests greedy nearest neighbor matching, in which each treated unit is sequentially matched with the k nearest control units. Matching depends on the ordering of the treated units, which you can specify in the ORDER= suboption.

You can specify the following suboptions:

K=k: specifies the number of matching control units, where k > 0, for each treated unit. PROC PSMATCH performs k separate loops of matching for treated units. In each loop, the nearest control unit is sequentially matched to each treated unit. By default, K=1 (one control unit for each treated unit).

Re: PSMATCH Procedure

sasphd — Mon, 20 Dec 2021 19:12:40 GMT

thanks for the response.

I wonder if there is a method in proc psmatch that give the optimal matching of control group?

Re: PSMATCH Procedure

Reeza — Mon, 20 Dec 2021 20:40:04 GMT

AFAIK matching is based on your experimental design not an optimization.