I'm trying to understand the hype around this estimation of heterogeneous treatment effects in the machine learning literature lately. It seems super interesting, but alot of it is beyond me. I read this paper by Susan Athey and I'm really struggling to understand how to implement this in clinical research. Can someone help explain this (with as simple non-technical terminology as possible) with an example of how to use this technique to uncover whether patients with certain characteristics who received a cancer drug were affected differently by the cancer drug than patients with different characteristics? Examples using code in R would be highly appreciated, but general examples would help too.
Asked
Active
Viewed 46 times
4
-
3A short technical paper distinguishing *effect modifiers* (variables correlated with effect heterogeneity) and *causal interactions* (variables that produce effects in combination that go beyond the sum of their solo contributions). Keele, L., & Stevenson, R. T. (2017). [Causal Interaction and Effect Modification: Same Model, Different Concepts](https://files.osf.io/v1/resources/azx52/providers/osfstorage/58a264bf594d9001f1ed7a92). – Alexis Nov 15 '19 at 04:51
-
@Alexis Thanks! Do you know of any good papers that are less technical? Also, I'm curious if someone could explain in practice how the forest models can do better with high dimensional data (large number of covariates) compared to the traditional regression model approaches – Jin Nov 15 '19 at 05:04
-
That's a separate question, and you should ask it. :) I recommend spending a little time with the Keele article. – Alexis Nov 15 '19 at 05:53