Parallelized prediction error estimation for evaluation of high-dimensional models

摘要: Summary: There is a multitude of new techniques that promise to extract predictive information in bioinformatics applications. It has been recognized first step for validation the resulting model fits should rely on proper use resampling techniques. However, this advice frequently not followed, potential reasons being difficulty correct implementation and computational demand. This addressed by R package peperr, which designed reliable prediction error estimation through resampling, potentially accelerated parallel execution compute cluster. Its interface allows easy connection newly developed fitting routines. Performance evaluation latter furthermore guided diagnostic plots, helps detect specific problems due high-dimensional data structures. Availability: http://cran.r-project.org, http://www.imbi.uni-freiburg.de/parallel Contact: cp@fdm.uni-freiburg.de Supplementary information:Supplementary are available at Bioinformatics online.