Standardized Mean Difference in PROC PSMATCH

wfan — Thu, 13 Jan 2022 19:42:00 GMT

I am puzzled by how PROC PSMATCH calculates the standardized mean differences for matched observations. According to the SAS document (https://documentation.sas.com/doc/en/statug/15.2/statug_psmatch_details25.htm), the standardized deviation for matched observation is pooled (before matching) instead of the matched (after matching). Why does SAS use the before-matching standardized deviation to calculate the after-matching standardized mean differences? This question also applies to weighting.

As an example, the table below is from the SAS document example (https://documentation.sas.com/doc/en/statug/15.2/statug_psmatch_examples01.htm). The same standard deviation is used for all observations and weighted observations. Shouldn't the standard deviations be different?

Many thanks

topic Standardized Mean Difference in PROC PSMATCH in Statistical Procedures

Standardized Mean Difference in PROC PSMATCH