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
        

Selection of Covariables

The selection of covariables is probably the most difficult problem in predictive modeling. Usually a large number of candidate predictors is available, which may have been studied before in similar or related diseases, or just have the interest of the investigator. In a data set of infinite size, we might test all predictors in univariable and multivariable analyses, and include the covariables with the highest statistical significance. In practice, data sets are often small relative to the number of candidate predictors. Therefore, more cautious selection methods should be used.

QUESTION 7.2

Some covariables may have strong correlations among each other. Choose which solution is to be preferred:

Selection ABecause of collinearity, a choice has to be made between one of the correlated variables.
Selection BBecause of no additional information, the most predictive of the related variables is to be included.
Selection CBecause of stability, the related variables should be combined.

Previous Page