search

bcallaway11 / did

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

Using xformla for not-yet-treated estimation #184

Open shrabasteebanerjee opened 8 months ago

shrabasteebanerjee commented 8 months ago

Hi,

I was wondering if it's possible to use time-invariant covariates if we have no control group and only not-yet-treated units? I have tried several times and it hasn't worked, despite no collinearity problems (local or global), so I was curious if I need to make any adjustments to use it in the not-yet-treated case? Thanks!

pedrohcgs commented 8 months ago

It is possible, but you must have covariates that do not fully predict treatment timing, and also you need the not-yet-treated group sample size needs to be bigger than the number of covariates.

If you have an unbalanced panel data, you also need all groups to be larger than the number of covariates in the model.

Thanks

On Fri, Oct 27, 2023 at 12:49 Shrabastee Banerjee @.***> wrote:

Hi,

I was wondering if it's possible to use time-invariant covariates if we have no control group and only not-yet-treated units? I have tried several times and it hasn't worked, despite no collinearity problems (local or global), so I was curious if I need to make any adjustments to use it in the not-yet-treated case? Thanks!

— Reply to this email directly, view it on GitHub https://github.com/bcallaway11/did/issues/184, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABE7347SWB4ZVQ3WKI7XPF3YBPQYXAVCNFSM6AAAAAA6TDC2OOVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE3DKOBWHA3TGMQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

paula-gomezcruz commented 8 months ago

Hi, I am having a similar issue, which I think might be cause by a lack of treated units in certain groups. Is there a way to exclude this groups from the analysis to be able to include the covariates and avoid this issue?

Thank you for help!