Selection Bias and Causal Inference in Empirical Studies of the U.S. Courts of Appeals