search

xuyiqing / fect

fixed effect counterfactual estimators
Other
51 stars 18 forks source link

method = both works differently for STATA and R packages #23

Open ahmedchaudhryy opened 3 years ago

ahmedchaudhryy commented 3 years ago

Hello,

Thank you for your amazing work with this package and the paper.

I have used the fect package on both R and STATA using the same dataset. When I specify method = "both" (while allowing cv to pick the best lambda or r) in either STATA or R, the model selection differs, on STATA it is picking up MC with lambda = 0 as the preferred model while on R its picking ife with r =3.

Can you please look into this matter? Which software's results should I trust more in this case?

Also a small request: Would it be possible to allow customization in the plot in R? In STATA we can choose the desired pre and post-treatment periods to be displayed on the plot, but it not exactly possible in R plot. We can choose proportion in R but it does not exactly work the same way as in choosing pre and off-periods. For example, I would want a plot with -14 pre-treatment periods and 6 post-treatment periods in R as well. (This would be really helpful in finalizing my estimates if R's results are to be trusted in model selection).

Thank you in advance!

xuyiqing commented 3 years ago

Would you mind sending this to my email account?

On Wed, Jun 30, 2021 at 2:04 AM Ahmed Chaudhry @.***> wrote:

Hello,

Thank you for your amazing work with this package and the paper.

I have used the fect package on both R and STATA using the same dataset. When I specify method = "both" (while allowing cv to pick the best lambda or r) in either STATA or R, the model selection differs, on STATA it is picking up MC with lambda = 0 as the preferred model while on R its picking ife with r =3.

Can you please look into this matter? Which software's results should I trust more in this case?

Also a small request: Would it be possible to allow customization in the plot in R? In STATA we can choose the desired pre and post-treatment periods to be displayed on the plot, but it not exactly possible in R plot. We can choose proportion in R but it does not exactly work the same way as in choosing pre and off-periods. For example, I would want a plot with -14 pre-treatment periods and 6 post-treatment periods in R as well. (This would be really helpful in finalizing my estimates if R's results are to be trusted in model selection).

Thank you in advance!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/xuyiqing/fect/issues/23, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB2PKGBFZYII5CFXKOIEC2TTVLMZRANCNFSM47R4JVZA .