search

insongkim / PanelMatch

113 stars 34 forks source link

How do results compare with other methods / Are my results significant? #45

Closed Mbjoerkh closed 1 year ago

Mbjoerkh commented 4 years ago

Hi again,

I've come across this issue a couple times where I appear to have found a significant result but where for the sake of simplicity, one data point is not significant. In the screenshot below, my eyeballing suggests there's a positive effect, on average, but obviously t+0 is not significant.

Is there a way to calculate the average effect (and corresponding standard errors) as opposed to the average effect at t0, t1, t2, and so on?

More generally, I think it would be very helpful if it is made easier to compare the results found with PanelMatch/PanelEstimate, with those of other more traditional methods such as 2WFE etc.

signi

kosukeimai commented 4 years ago

Thanks for the suggestion. It's a little tricky because they are all correlated, but we will add this to the to-do list.

Mbjoerkh commented 4 years ago

Thanks.

It would seem to me that one might be able to use other existing estimation methods/packages on the matched sets from PanelMatch, but maybe I underestimate the standard error calculations. In any case, I would think this will be a common crossroad for users as reviewers will want to know.

Thanks again!

adamrauh commented 1 year ago

Closing this issue as I believe it has been implemented -- discussed over in #99