JSM 2014 Home
Online Program Home
My Program

Abstract Details

Activity Number: 244
Type: Contributed
Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistics in Epidemiology
Abstract #313625
Title: Coupling Model Selection with Evaluation in Prognostic Model Development
Author(s): Aasthaa Bansal*+ and Patrick Heagerty
Companies: University of Washington and University of Washington
Keywords: Risk prediction ; Prognosis ; Model selection ; Biomarkers ; Medical decision making
Abstract:

Prognostic models can be used to prioritize medical treatments by identifying patients who are at greatest risk of event. It is common practice to build a statistical model combining several variables, with the goal of obtaining a risk score with improved performance. This involves model selection, followed by evaluation. Recent literature has focused on model evaluation, which may be achieved using one of many measures such as the c-statistic (or area under the ROC curve) or the net benefit, depending on the scientific application. Regardless of the metric used for evaluation, model selection typically first takes place by optimizing standard goodness of fit measures such as Akaike's Information Criterion. This results in a disconnect between model search and evaluation, as the best fitting model does not necessarily translate to being the most predictive model. We propose an approach that couples model selection with evaluation, so that the metric of interest by which we ultimately evaluate model performance is optimized in the initial model selection stage. We illustrate these ideas and our proposed approach using a dataset from a multiple myeloma study of candidate biomarkers.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2014 program




2014 JSM Online Program Home

For information, contact jsm@amstat.org or phone (888) 231-3473.

If you have questions about the Professional Development program, please contact the Education Department.

The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.

ASA Meetings Department  •  732 North Washington Street, Alexandria, VA 22314  •  (703) 684-1221  •  meetings@amstat.org
Copyright © American Statistical Association.