seems to me they used the same data that was used to fit the models, to estimate prediction accuracy. This overestimates prediction accuracy. It would be better

seems to me they used the same data that was used to fit the models, to estimate prediction accuracy. This overestimates prediction accuracy. It would be better to use data that was not used to fit the models to assess accuracy. This is known as cross validation