Hi there,
I'm performing a propensity score matching exercise, whereby I obtain an average treatment effect (ATE). To obtain the ATE, untreated subjects are matched on propensity scores to treated subjects are and vice versa to see how their outcomes would differ. I would like to compute some sort of diagnostic for my propensity score model.
I've read a lot of documentation regarding the computation of effect sizes for all covariates both before and after a matched sample has been obtained. However, this only seems to be performed when obtaining the average treatment effect on the treated (ATET) or the average treatment effect on the untreated (ATEU), whereby either the treated OR the untreated subjects are matched to their counter-parts, not both - as in my case.
I don't see how this diagnostic can be performed when matches are performed in both directions.
If indeed it can, could someone point me in the right direction. If it cannot, has anyone had any experience of computing diagnostics for this propensity score matching procedure?
Thanks.
I'm performing a propensity score matching exercise, whereby I obtain an average treatment effect (ATE). To obtain the ATE, untreated subjects are matched on propensity scores to treated subjects are and vice versa to see how their outcomes would differ. I would like to compute some sort of diagnostic for my propensity score model.
I've read a lot of documentation regarding the computation of effect sizes for all covariates both before and after a matched sample has been obtained. However, this only seems to be performed when obtaining the average treatment effect on the treated (ATET) or the average treatment effect on the untreated (ATEU), whereby either the treated OR the untreated subjects are matched to their counter-parts, not both - as in my case.
I don't see how this diagnostic can be performed when matches are performed in both directions.
If indeed it can, could someone point me in the right direction. If it cannot, has anyone had any experience of computing diagnostics for this propensity score matching procedure?
Thanks.
