A model that can accurately predict a person’s risk for developing a particular disease is, obviously, extremely useful. For one thing, the ability to associate risk factors with disease likelihood provides the opportunity to take preventative measures that can save lives.
The goal of modeling risk of disease is therefore to identify potential prediction variables and develop a model that can be clinically useful. Typically, risk of disease is modeled with a binary response variable using logistic regression. But other modeling methods can be used as well. How do they compare?
In this webinar, Andrea Coombs walks through an example based on building predictive models for peripheral arterial disease risk using data from the National Health and Nutrition Examination Surveys. She demonstrates how to:
- Build traditional logistic models.
- Create alternative models using more modern modeling approaches.
- Apply validation and model comparison techniques to determine the best model for disease risk prediction.