post-estimation command that does a multiverse analyses to assess incidences where common support is lacking

mhamjediers / nopo_decomposition

Stata ado files to run matching-based decompositions

MIT License

3 stars 0 forks source link

post-estimation command that does a multiverse analyses to assess incidences where common support is lacking #19

Closed mhamjediers closed 4 months ago

mhamjediers commented 5 months ago

First draft is done nopo commsupport; still some minor things that are open:

[x] adjust all local/variable names with underscores
[x] carry forward weights --> no, only a note that share of matched are based on unweighted obs
[x] add all options for the twoway-graphs (also one option for markers for used/unsued variables)
[x] add in help-file
[x] check whether varlist in option is part of the matchset
[x] check whether it works with linebreaks in variable-labels
[x] add tuples to the requirements (alternatively use the small chunk of their ado-file and acknowledge it; it's really short)

mhamjediers commented 4 months ago

@maximilian-sprengholz : in the newest version, I also adjusted the return _matchset of nopo decomp (lines 334-348)

Do we need to omit potential multiples (and/or interaction)? Otherwise check also a shorter version of the code in the line (350-356).

maximilian-sprengholz commented 4 months ago

I implemented the doublettes check because all predictors are returned 3 times for kmatch em: in e(xvars) , e(ematch) and e(emxvars). Would be good for commsupport to have a unique list, or do you check again (then I would prefer ensuring uniqueness in the ereturns already).

Elegant solution with fvrevar - didn't know that one.