A note on out-of-sample prediction, marginal effects computations, and temporal testing with random parameters crash-injury severity models