Closed mhamjediers closed 1 year ago
The wor
excemption makes sense to me, but everything else should be accomodated by the weights, no? Default in kmatch md
is kernel matching with automatically determined bandwidth. Nearest neighbor matching has to be specifically requested via nn(#)
.
Does now abort after kmatch md ..., nn() wor
:
// abort if matched without replacement (replacement necessary for correct weights)
if ("`e(wor)'" == "wor") {
dis as error "nopo decomp requires matching with replacement. Option wor not allowed in:"
dis as error " `_kmatch_cmdline'"
error 322
exit
}
Closing.
If yes, the kmatch-postestimation should have a check and error message if not nn(1) or if wor (without replacement) specified