Homepage of Justin Tobias, Assistant Professor of Economics, UCI

Below are results obtained using an alternate prior. In this prior, we keep all the hyperparameters the same as those presented in the paper, except for the correlation between the observed and potential outcomes which is now centered at .5.

All of the tables and figures are provided in .pdf format.

Table of Coefficient Posterior Means, Standard Errors, and probabilities of being positive.
Graph of Non-Identified Correlation Parameter and its Upper and Lower Conditinal Support Bounds
Graph of ATE and TT posterior predictives.

For the ORIGINAL PRIOR used in the paper, here is a graph of ATE and TT for the first component of the mixture only. Note that this suggests that for a ``minority'' subpobulation, selection does not seem to be an issue, though as a whole, selection bias is a key feature of this data. Graph of ATE and TT First Component of Mixture Only.