American Statistical Association
Calibration is an important measure of the predictive accuracy of a model. A widely used measure of calibration when the outcome is binary or survival time is the expected Brier score; a mean squared error loss function. The objective in this talk is to evaluate the impact of a subset of biomarkers on the Brier score. When the endpoint is survival time, survival functions from nested proportional hazards models are frequently used to assess the improvement in predictive accuracy due to the additional markers. An underappreciated problem with this approach is that nested proportional hazards models cannot jointly be true. An alternative approach, projecting the full model survival function onto a lower dimensional space, representing the survival function without the biomarkers of interest, is proposed. A prostate cancer data set is used to illustrate the method.
|Date:||Wednesday, February 14, 2018|
|Time:||4:00 - 5:00 P.M.|
Memorial Sloan Kettering Cancer Center
Department of Epidemiology and Biostatistics
485 Lexington Avenue
(Between 46th & 47th Streets)
2nd Floor, Conference Room B
New York, New York
**Outside visitors please email firstname.lastname@example.org for building access. You must be on the security list to enter the floor.