Hello,
I am quite new to causal inference so I am very sorry if the question will sound trivial
I have a panel dataset of 50 countries over 18 years. Whereby some have received treatment and others did not.
However 1) treatment occur at different times for different countries, (i.e. some country have the treatment at time 0 some at time 2 some at time 5 etc) 2) treatment is not a dummy, but rather a continuous variable. (I.e some have a treatment of 1 other have a treatment of 3 etc.) 3) some countries receive the treatment multiple times.
I was wondering whether it was possible to estimate such dataset as with a diff and diff. I do get the concept for 2 groups, but I am not sure in practice how I could apply the same concept to panel data. Also, I would like to avoid wasting the information on the value of the treatment if possible.
Is this a usual DIff and diff context in panel data? is it possible to estimate it? do you have suggestions on how to go about it?
It would be particularly useful to have some indication of the code of how to implement this. In particular, how would I define when a country is part of the control or the treated group, also when a country would be in the pre/post-treatment?
I thank you in advance for your help
Best regards
I am quite new to causal inference so I am very sorry if the question will sound trivial
I have a panel dataset of 50 countries over 18 years. Whereby some have received treatment and others did not.
However 1) treatment occur at different times for different countries, (i.e. some country have the treatment at time 0 some at time 2 some at time 5 etc) 2) treatment is not a dummy, but rather a continuous variable. (I.e some have a treatment of 1 other have a treatment of 3 etc.) 3) some countries receive the treatment multiple times.
I was wondering whether it was possible to estimate such dataset as with a diff and diff. I do get the concept for 2 groups, but I am not sure in practice how I could apply the same concept to panel data. Also, I would like to avoid wasting the information on the value of the treatment if possible.
Is this a usual DIff and diff context in panel data? is it possible to estimate it? do you have suggestions on how to go about it?
It would be particularly useful to have some indication of the code of how to implement this. In particular, how would I define when a country is part of the control or the treated group, also when a country would be in the pre/post-treatment?
I thank you in advance for your help
Best regards
