xthdidregress: Aggregating ATETs following estimation

[Kristoffer Madsen]

Hi Statalist,

Having estimated each ATET(g,t) using xthdidregress, several postestimation commands for aggregation are available via -estat-, including the -overall- and -dynamic- options.

Has anyone managed to easily compute an "overall" ATET estimate with corresponding standard errors and confidence intervals ONLY for selected ATET(g,t)? For example, one might have 8 pre-treatment periods and 8 post-treatment periods, but only want to estimate an ATET for -4 to +4. Might there be an easy way to achieve this rather than slicing the underlying data to be restricted to the wanted number of pre and post-treatment periods?

[FernandoRios]

Two options
1. re estimate, constraining the sample to your specific periods (may be hard if you are using not yet treated)
2. Use csdid2. It has options for exactly that

[FernandoRios]

Otherwise. use the event option, and lincom (with some weights) for custom aggregation

[Kristoffer Madsen]

Two options
1. re estimate, constraining the sample to your specific periods (may be hard if you are using not yet treated)
2. Use csdid2. It has options for exactly that

1) Exactly - it quickly becomes cumbersome.
2) Huge fan of csdid, but haven't tried csdid2. I will give it a shot.

[Kristoffer Madsen]

Otherwise. use the event option, and lincom (with some weights) for custom aggregation

Thanks for the suggestion, Fernando. Hadn't thought of lincom. However, using csdid2 is probably more robust to user errors, so I'll start with that.