... so that it automatically invokes marginaleffects::comparisons under the hood for emfx. (i.e. evaluates mean(.Dtreat==1 - mean(.Dtreat==0)instead ofmean(dYdX)`).
This won't make a difference for linear models, but will slightly change the results for non-linear families. So it's a bug fix.
Ironically, this was the default in the first iterations of the package. But I changed it whilst troubleshooting the jwdid difference in #4.
... so that it automatically invokes
marginaleffects::comparisons
under the hood foremfx
. (i.e. evaluatesmean(.Dtreat==1
- mean(.Dtreat==0)instead of
mean(dYdX)`).This won't make a difference for linear models, but will slightly change the results for non-linear families. So it's a bug fix.
Ironically, this was the default in the first iterations of the package. But I changed it whilst troubleshooting the
jwdid
difference in #4.