Discovering Latent Treatments in Text Corpora and Estimating
Their Causal Effects
Description
Implements the approach described in Fong and Grimmer (2016) for
automatically discovering latent treatments from a corpus and estimating the average marginal component effect (AMCE) of
each treatment. The data is divided into a training and test set. The supervised Indian Buffet Process (sibp) is used
to discover latent treatments in the training set. The fitted model is then applied to the test set to infer the values
of the latent treatments in the test set. Finally, Y is regressed on the latent treatments in the test set to estimate
the causal effect of each treatment.