Statistical Decision-Making Models and Treatment Effects

Abstract:     
Statistical models are better at prediction than intuitive judgment. However, in certain situations an expert's judgment may have a large effect on outcomes, for example expert witnesses testifying in court. Building statistical models from outcome data affected by such judgments requires adjustments for treatment effects. We present an example of how to model and bound such treatment effects when data from the comparison group is not available.

Number of Pages in PDF File: 23

Keywords: Treatment effects, actuarial models, non-experimental methods, bounding analysis

JEL Classification: C15, C82, C99

working papers series