LARS software for R and Splus

Authors: Brad Efron and Trevor Hastie
Maintainer: Trevor Hastie

The software, written in the S language, computes the entire LAR, Lasso, or (epsilon) forward stagewise coefficient path in the same order of computations as a single least-squares fit. For slightly more details, see announce.txt

  • LARS in R: a fully documented and maintained R package called lars: "lars" from the contributed packages section of the R website
    (via CRAN). These packages include the diabetes data used in the LARS paper.

  • LARS in Splus (not updated or maintained): functions and Splus helpfiles (a shell archive).

  • The diabetes data (tab delimited, and as depicted on page 2) used in the paper. Note that these data are first standardized to have zero mean and unit L2 norm before they are used in the examples. Here are the standardized data, and the 442 x 64 matrix for the "quadratic model"

    [Home] back to Trevor Hastie's Software