search

zachporreca / staggered_adoption_synthdid

NO LONGER MAINTAINED. Links to alternative packages are in the readme. Code to incorporate staggered treatment adoption (based on appendix from Arkhangelsky et al. 2021) into synthdid package
21 stars 10 forks source link

Variance estimator #4

Closed regulyagoston closed 1 year ago

regulyagoston commented 1 year ago

Hi Zachary,

I really appreciate your function implementing staggered synthetic diff-in-diff!

I started to use for my empirical problem and when going through the codes and your paper, I see a potential difference: in the codes:

line 92: variance=t(weights) %*% weights.

I believe it should be: variance=t(weights) %% variance %% weights, according to equation 16 from your paper. (Otherwise influence calculation is not used.)

Can you confirm?

Many thanks, Agoston

zachporreca commented 1 year ago

Wow! Great catch! Thanks so much

zachporreca commented 1 year ago

This was a good opportunity, to switch to the placebo algorithm suggested in the new Athey paper. I genuinely appreciate your insight here!

regulyagoston commented 1 year ago

Hi,

You have the clustering paper in mind by Athey, Imbens and Wooldridge? That would be truly useful as with my finance application I would like to incorporate that effect somehow. Btw do you know the paper Uncertainty Quantification in Synthetic Controls with Staggered Treatment Adoption by Cattaneo et al? ( https://mdcattaneo.github.io/papers/Cattaneo-Feng-Palomba-Titiunik_2022_wp.pdf) would be interested in your thoughts how it connects to your work. As far as I see they propose a non-asymptotic method. The package is quite tedious and not ready yet, but seems interesting to me.

Bests, Agoston

On Wed, Apr 5, 2023, 13:54 Zachary Porreca @.***> wrote:

This was a good opportunity, to switch to the Bootstrap algorithm suggested in the new Athey paper. I genuinely appreciate your insight here!

— Reply to this email directly, view it on GitHub https://github.com/zachporreca/staggered_adoption_synthdid/issues/4#issuecomment-1497890874, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJT4GGWUWGZ2ZXSZUT7X4F3W7WWTPANCNFSM6AAAAAAWUBZMNU . You are receiving this because you authored the thread.Message ID: @.***>

zachporreca commented 1 year ago

No, it's this Clarke et al. (2023) paper Synthetic Difference-in-Differences Estimation (I have it linked on the repository home page now).....I've coded up their placebo and jackknife methods now.

As for the uncertainty quantification paper, I did see that presented at a workshop over the winter and got to hang out a bit with one of the authors. Really brilliant guy.....It's a great paper and method that I am sure will make some waves, but for the time being a bit above my paygrade!

regulyagoston commented 1 year ago

I see, that is a nice and straight forward suggestion! Thanks for extending the code! I will go through it!

On Apr 6, 2023, at 6:54 AM, Zachary Porreca @.***> wrote:

No, it's this Clarke et al. (2023) paper Synthetic Difference-in-Differences Estimation (I have it linked on the repository home page now).....I've coded up their placebo and jackknife methods now.

As for the uncertainty quantification paper, I did see that presented at a workshop over the winter and got to hang out a bit with one of the authors. Really brilliant guy.....It's a great paper and method that I am sure will make some waves, but for the time being a bit above my paygrade!

— Reply to this email directly, view it on GitHub https://github.com/zachporreca/staggered_adoption_synthdid/issues/4#issuecomment-1498877436, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJT4GGXXCCNT2M3MRR2TQX3W72OH3ANCNFSM6AAAAAAWUBZMNU. You are receiving this because you authored the thread.