Update README - Githubissues

All your suggested changes look correct to me, thanks!

On Tue, Jun 20, 2023, 9:38 PM Mauricio Caceres Bravo < @.***> wrote:

@.**** commented on this pull request.

Some minor edits. LMK if you want me to make them.

In README.md https://github.com/mcaceresb/stata-pretrends/pull/2#discussion_r1236176829 :
+In this case, these coefficients come from a two-way fixed effects regression using reghdfe. However, the pretrends package will work with any package that produces an event-study from any +asymptotically normal estimator, including +Callaway and Sant’Anna (2020) +and Sun and Abraham (2020), so long as the resulting estimates and coefficients are saved in e(b) and e(V). If one is using a command that does not export an e(b) and e(V), one can instead provide the coefficients and covariance matrix directly via the beta() and sigma() options. +

The option is beta() and vcov() (alias b() and v()).

In README.md https://github.com/mcaceresb/stata-pretrends/pull/2#discussion_r1236177393 :
r(Power) = .5
-To visualize the linear trend against which pre-tests have 50 percent power: +In the command above, the option pre(1/3) tells the package that the pre-treatment event-study coefficients are in positions 1 through 3 in our regression results. (The package assumes that the period before the event-study is normalized to zero and omitted from the regression.) Likewise, the option post(1/4) tells the package that the post-treatment coefficients are in positions 4 through 7. The results of the command tells us that if there was a linear violation of parallel trends with slope 0.049, then we would have 50% power to detect it (where we say it's "detected" if there is a significant pre-trend coefficient). If we want wanted a different power threshold, say 80%, we would change power 0.5 to power 0.8 in the command above.

It would be post(4/7)

In README.md https://github.com/mcaceresb/stata-pretrends/pull/2#discussion_r1236178670 :
r(Power) = .5
-To visualize the linear trend against which pre-tests have 50 percent power: +In the command above, the option pre(1/3) tells the package that the pre-treatment event-study coefficients are in positions 1 through 3 in our regression results. (The package assumes that the period before the event-study is normalized to zero and omitted from the regression.) Likewise, the option post(1/4) tells the package that the post-treatment coefficients are in positions 4 through 7. The results of the command tells us that if there was a linear violation of parallel trends with slope 0.049, then we would have 50% power to detect it (where we say it's "detected" if there is a significant pre-trend coefficient). If we want wanted a different power threshold, say 80%, we would change power 0.5 to power 0.8 in the command above. + + +Next, we illustrate how to visualize violations of parallel trends using the package's second subcommand. For simplicitly, lets visualize the linear trend against which pre-tests have 50 percent power that we just calculated. This is just for illustration --- you can visualize any violation that you want, and should choose an economically relevant one. To do this, we run the command:

I believe --- is replaced with a horizontal line, rather than an em-dash. You can insert an em-dash directly, —, or via HTML, —.

In README.md https://github.com/mcaceresb/stata-pretrends/pull/2#discussion_r1236179577 :

return list

scalars: - r(LR) = .4332635208743188 - r(Bayes) = .3841447004284795 - r(Power) = .6624492444726371 - r(slope) = . +* r(LR) = .3962795305253372

I believe this would show up misaligned?

— Reply to this email directly, view it on GitHub https://github.com/mcaceresb/stata-pretrends/pull/2#pullrequestreview-1489482844, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE6EXFAYIUYNF2Q7P3AOHOTXMJGAJANCNFSM6AAAAAAZOAMO4E . You are receiving this because you authored the thread.Message ID: @.***>

mcaceresb / stata-pretrends

Update README #2

Some minor edits. LMK if you want me to make them.

The option is beta() and vcov() (alias b() and v()).

It would be post(4/7)

I believe --- is replaced with a horizontal line, rather than an em-dash. You can insert an em-dash directly, —, or via HTML, —.