Model selection

  • Please report on the fit of your model

    • E.g., R2, Adj. R2, F test (for regressions)
    • E.g., Log likelihood test against a null model, hit rate, hit probabilities (for Logit models)
    • Validation on a holdout sample
  • If you have estimated “competing” models, show the fit statistics for all models, and then motivate your final model choice.

  • Provide diagnostic plots, where appropriate.


    Source: Datta, H., Foubert, B., & Van Heerde, H. J. (2015). The challenge of retaining customers acquired with free trials. Journal of Marketing Research, 52(2), 217 234.