Personalized Predictions Using Exogenous Covariates

• Activity Number: 503
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Jie Fan* and J. Sunil Rao
• Companies: and University of Miami
• Keywords: The mixed model ; classified mixed prediction ; exogenous covariates ; personalized predictions

Abstract

Personalized predictions are a holy grail of medical research. These predictions can be estimated by a technique known as classified mixed prediction (CMP) via a linear mixed model by first correctly classifying the random effect associated with the new observation. However, when exogenous covariates are present, as is the case when making predictions on members of the population from which the mixed model was not derived, things get much trickier. We describe a series of different prediction paradigms under this umbrella, and present systematic solutions for each. We demonstrate how to make personalized predictions for survival time for individuals who have colorectal cancer and who are members of minority subgroups. The exogenous covariates are significantly different hypermethylated sites across minority subgroups.