Fridge: Focused fine-tuning of ridge regression for personalized predictions.

摘要: Statistical prediction methods typically require some form of fine-tuning tuning parameter(s), with K-fold cross-validation as the canonical procedure. For ridge regression, there exist numerous procedures, but common for all, including cross-validation, is that one single parameter chosen all future predictions. We propose instead to calculate a unique each individual which we wish predict an outcome. This generates individualized by focusing on vector covariates specific individual. The focused ridge-fridge-procedure introduced 2-part contribution: First define oracle minimizing mean squared error covariate vector, and then estimate this using plug-in estimates regression coefficients variance parameter. procedure extended logistic parametric bootstrap. high-dimensional data, use estimate, simulations show fridge gives smaller average than both simulated real data. illustrate new concept linear models in 2 applications personalized medicine: predicting risk treatment response based gene expression method implemented R package fridge.