Skip to Content
Interactive Textbook on Clinical Symptom Research Logo


Home Button

Statistical Models for Prognostication
Author Bio
Introduction
Predictions: Statistical Models
Insight: Statistical Models
Ingredients: Statistical Models
Theoretical Aspects
Central Concepts
Currently selected section: Regression Models
Problems: Regression
Practical Advice
Example 1
Example 2
Chapter 8: Statistical Models for Prognostication: Development of Regression Models
        

Validation

Validation is the process of determining the validity of a model in new patients, and is an important aspect of model development. It makes no sense to develop a model that nicely describes the patterns in the data set under study, but fails when tested in new patients. Validity may be distinguished in internal and external validity (Justice et al., 1999) (Altman and Royston, 2000).

QUESTION 7.7

Internal validity refers to the validity of the model:

Selection AAccording to clinical experts.
Selection BIn the sample that is used to construct the model.
Selection CIn the population where the sample originated.
Selection DIn similar but slightly different populations.

We next look at:

  • what model aspects to validate, and
  • techniques for validation.

Aspects to Validate

In principle, all aspects of model development should be validated. This includes:

  • the coding and selection of predictors
  • the estimation of regression coefficients, and
  • the evaluation of model performance.

When the whole process of model development is considered, the validity of the final model and the quality of its predictions can honestly be assessed. This is in contrast to the situation in which, for example, the model is considered as pre-specified while in fact data-driven decisions were made to specify the model (especially the selection of predictors). Then, model performance would be overestimated.

 

Previous Page