dajmcdon/cplr: Compressed and penalized linear regression

This package implements the simulations and data analysis examples for compressed and penalized linear regression as in (Homrighausen and McDonald, 2017). Essentially, the design matrix is premultiplied by a sparse matrix, reducing the number of available observations from n to q. However, the addition of a ridge penalty results in estimates of the true coefficient vector with lower mean-squared error, even relative to ridge regression (in some cases). The result is improved computation and better statistical accuracy.

install.packages("remotes")
remotes::install_github("dajmcdon/cplr")