Keith Battocchi

For the first question, this is saying setting race from Amer-Indian-Eskimo to Black on average would reduce the output of the classifier by 1.8 (e.g. for a particular datapoint this...

I don't think I'd say the assumptions on the DGP are less restrictive for DeepIV, they're just different. The functional form of the relationship between Y and T can be...

Yes, although primarily designed for discrete treatments, SingleTreePolicyInterpreter can also be used with a continuous treatment; in that case the policy will effectively tell you if the marginal value of...

Most of our estimators are assuming a linear treatment effect, so if marginal effect is positive (which is what the tree is showing) then continuing to increase the treatment intensity...

EconML's CATE estimators are based around telling you how an outcome (Y) will _change_ given some change in the treatment (T), conditional on some confounders (X). If you have historical...

This part is not quite right: > counter_factual_output = historical_output - est.effect(X=features, T0=original_treatment, T1=T0*X% decrease)). because you still ought to add the effect, not subtract it (but if decreasing the...

You're right that if X can perfectly predict T then there is no residual that we can use to predict the treatment effect when we're using estimators like DML, so...

Could you provide a little bit more context on what method you called and what the arguments were to produce that output?

Most of our estimators assume a linear relationship between the treatment(s) and outcome(s), conditional on the features. As you note, this may not be realistic, particularly over a wide range...

Yes, that estimator assumes a constant marginal treatment effect, so the total effect will just be the marginal effect times the change in treatment (25 * (-2.7) in your case)....