csdid coefficients

[Bo Hou]

Hi

Long term follower of this forum. First time to post.

I am working on an evaluation project of a health programmer using csdid. I found difficult to interpret the estimated ATT coefficient.

My outcome is a dummy variable (hospitalization or not), the coefficient was say -0.02. Is this probability - the same as Linear Probability Model? Also is it possible to show odds ratios instead rather than coefficient? What about count variables? If so, how?

Regards
Bo

[FernandoRios]

You are correct
yhe coefficient is similar to the interaction in a linear probability model
thus, no, you can’t get odds ratios

[Jeff Wooldridge]

My 2023 Econometrics Journal paper shows how to use logit and Poisson regression. I need to know more. Is this panel data with staggering?

[FernandoRios]

True
if you change csdid with jwdid
And add the option method(logit)
you will get the probability which could be transformed into odds ratios

[Bo Hou]

My 2023 Econometrics Journal paper shows how to use logit and Poisson regression. I need to know more. Is this panel data with staggering?

Thanks you so much for replying to this. I am a fan of your textbook. My econometrics is basic. I will try to give more details.

We have 20 quarters of data of 5000 patients. Treatment groups started at different time say from quarter 7 to 14, then quarter 16-20 is follow up period. It is a social prescribing intervention- so we assume if someone is on the program then they are always on. Thus, I think this might be a panel data with staggering.

Would this work?

[Bo Hou]

True
if you change csdid with jwdid
And add the option method(logit)
you will get the probability which could be transformed into odds ratios

Thanks very much. I will take a look.