Erratum in: Journal of Causal Inference 2014;2(1):109–12, DOI: 10.1515/jci-2014-0017
Line 5 of Theorem 1 should read:
2(ii). is identified assuming a model where Z is randomized.
Third sentence after Remark should read:
Condition 2(ii) requires nonparametric identification of the average treatment effects (ATE) [4–6] in a modified model, with Z fixed. This identification can be readily decided by graphical means .
Line 5 of Theorem 2 should read:
2(ii). is identified assuming a model where X is randomized.