CME with continuous predictor

https://academic.oup.com/ectj/article-abstract/24/2/264/5899048?redirectedFrom=fulltext

Conditional Marginal Effect (Slope) with continuous Z: E[dY/dD|Z=z]. Two step: (1) fit (2) reduce to 1d with bins or kernel. Yiqing Xu points to Semenova & Chernozhukov