search

bcallaway11 / did

Difference in Differences with Multiple Periods, website: https://bcallaway11.github.io/did
287 stars 91 forks source link

Interpretation of 'average effect by length of exposure' figure #146

Closed cjing1 closed 1 year ago

cjing1 commented 1 year ago

Thank you again for the great package. I have another question regarding the interpretation of the 'average effect by the length of exposure' figure. I learned that we could check if the confidence intervals cover 0 in all pre-treatment periods to test the parallel assumption.

However, in my case, the time unit -1 does not cover 0, but all other pre-trt periods do.

I know that -1 corresponds to the time period before groups first participate in the treatment. But I am not sure how to interpret the result since it is "before" exposure. Is this an indicator of violation of parallel trend? If so, is there a way to solve it? Maybe matching on trt-control?

Thank you in advance for any help or suggestions.

pedrohcgs commented 1 year ago

This may be a violation of the no-anticipation assumption. You may allow for 2 periods of anticipation if you wish.

Without knowing more about the details, it is hard to say something concrete. So take it with a grain of salt.

On Tue, Sep 27, 2022 at 20:56 ChenJ @.***> wrote:

Thank you again for the great package. I have another question regarding the interpretation of the 'average effect by the length of exposure' figure. I learned that we could check if the confidence intervals cover 0 in all pre-treatment periods to test the parallel assumption.

However, in my case, the time unit -1 does not cover 0, but all other pre-trt periods do.

[image: image] https://user-images.githubusercontent.com/31779053/192667956-d8306d14-f07d-4460-bebf-35cb928bdae0.png

I know that -1 corresponds to the time period before groups first participate in the treatment. But I am not sure how to interpret the result since it is "before" exposure. Is this an indicator of violation of parallel trend? If so, is there a way to solve it? Maybe matching on trt-control?

Thank you in advance for any help or suggestions.

— Reply to this email directly, view it on GitHub https://github.com/bcallaway11/did/issues/146, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABE7342XBVRF5YWU4YUA57TWAOQTRANCNFSM6AAAAAAQXKDNXI . You are receiving this because you are subscribed to this thread.Message ID: @.***>

--

Pedro H. C. Sant'Anna Department of Economics Vanderbilt University 615-875-8448 (phone) @.*** https://pedrohcgs.github.io

cjing1 commented 1 year ago

Thank you Prof. Sant'Anna for such quick responses. I am studying the effect of news coverage on public opinion. So I think you are right about the anticipation issue. This is my first time dealing with it, sorry if my question is basic. May I ask how to allow for 2 periods of anticipation? Should I extend the beginning of treatment to a one-time unit before?

Thank you so much for your help.

bcallaway11 commented 1 year ago

The att_gt function has an anticipation argument which you can use to allow for anticipation directly.