Model assumptions are important for the quality of predictions from a regression model. Better predictions will result from a model that satisfies its underlying assumptions. However, assumptions can never be fully met in empirical data, and evaluation preferably focuses on the validity of the predictions in new patient data.