multiple treatments confounding

[jbel1026]

@vsyrgkanis for multiple treatments, is it valid to use Treatment A in the propensity model for Treatment B?

for example would we want to estimate f1 as E[T1|X, W T2]? if T2 is a potential confounder of T1 and Y, I dont think the econML package does this,is this a functionality that could be added? Is there a reason it wasnt, that I am not understanding?

My fear is that if we dont do this we could have co-linearity between the treatment effects in the final stage model

https://us-prod.asyncgw.teams.microsoft.com/v1/objects/0-eus-d9-3c39d23fd75ac3112d217460324ece37/views/imgo