causal inference

[gustavolacerda]

Although I love graphical models, I suspect that the "potential outcomes approach" is a more proper treatment of causal inference. See new book by Hernan and Robbins (h/t Michael Sobel).

Key concepts:
* counterfactuals: should be well-defined, but often aren't in the social science literature (e.g. to answer "what is the effect of marriage on

Comments

[random_walker]

i seem to recall having said this earlier.

[gustavolacerda]

which part?