Empirical Analysis of Model Selection for Heterogenous Causal Effect Estimation