Skip to main content
eScholarship
Open Access Publications from the University of California

UC Berkeley

UC Berkeley Electronic Theses and Dissertations bannerUC Berkeley

Software for prediction and estimation with applications to high-dimensional genomic and epidemiologic data

  • Author(s): Ritter, Stephan Johannes
  • Advisor(s): Hubbard, Alan E.
  • et al.
Abstract

Three add-on packages for the R statistical programming environment (R Core Team, 2013) are described, with simulations demonstrating performance gains and applications to real data. Chapter 1 describes the relaxnet package, which extends the glmnet package with relaxation (as in the relaxed lasso of Meinshausen, 2007). Chapter 2 describes the widenet package, which extends relaxnet with polynomial basis expansions. Chapter 3 describes the multiPIM package, which takes a causal inference approach to variable importance analysis. Section 3.7 describes an analysis of data from the PRospective Observational Multicenter Major Trauma Transfusion (PROMMTT) study (Rahbar et al., 2012; Hubbard et al., 2013), for which the multiPIM package is used in conjunction with the relaxnet and widenet packages to estimate variable importances.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View