Work

xrnet: Hierarchical Regularized Regression to Incorporate External Data

Year: 2019

Type: article

Abstract: Regularized regression is an essential tool for both feature selection and prediction with highdimensional data.A number of R (R Core Team, 2019) packages have been developed to fit regularized regression models, including glmnet (Friedman, Hastie, & Tibshirani, 2010), biglasso (Zeng & Breheny, 2017), and ncvreg (Breheny & Huang, 2011).These packages can fit multiple model types including linear, multivariate linear, logistic, and Cox regression with different regularization penalties.The penalties control the complexity of the models by shrinking the coefficients toward zero, with the degree of shrinkage controlled by a tuning parameter that is typically selected by cross-validation.In addition to shrinkage, penalties like the lasso, elastic-net, SCAD (Fan & Li, 2001), and MCP (Zhang, 2010) also perform feature selection by shrinking some coefficients to exactly zero.

Source: The Journal of Open Source Software

DOI: 10.21105/joss.01761

Authors: Garrett Weaver, Juan Lewinger

Institution: University of Southern California

Language: English

Topic: Statistical Methods and Inference

Subfield: Statistics and Probability

Field: Mathematics

Domain: Physical Sciences

Open Access status: diamond

Funders: National Institutes of Health, National Cancer Institute

Award: Statistical Methods for Integrative Genomics in Cancer