Entropy weights

[Adrienne Wold]

Hi all,

I am using the ssc installed command ebalance for entropy balancing. I am conducting a test measuring differences among survey respondents in an outcome before and after events. In addition to my treatment variable, I have event FEs and district FEs.

I am wondering if this is the correct way to create the weights:

ebalance treat covars , gen(eb_weights)

Or whether I also include the event FEs and district FES and additional covariates when creating the weights.

[Felix Bittmann]

A newer implementation is ebalfit, see: https://www.statalist.org/forums/for...ropy-balancing

If each cluster is large enough, you can attempt to include it as an FE. If there are too many and smaller clusters present, the balancing might fail. If so, using the simpler model also appears fine.