In MGallow/logitr: Penalized Logistic Regression

knitr::opts_chunk$set(echo = TRUE, cache = TRUE)

***This package is experimental and not regularly maintained. It is recommended to use {glmnet} instead!

Overview

logitr is an R package for linear and logistic regression with optional ridge and bridge regularization penalties. A (possibly incomplete) list of functions contained in the package can be found below:

logisticr() computes the coefficient estimates for logistic regression (ridge and bridge regularization optional)
linearr() computes the linear regression coefficient estimates (ridge regularization and weights optional)
predict.logisticr() generates predictions and loss metrics for logistic regression
predict.linearr() generates predictions and loss metric for linear regression

See vignette or manual.

Installation

# The easiest way to install is from the development version from GitHub:
# install.packages("devtools")
devtools::install_github("MGallow/logitr")

If there are any issues/bugs, please let me know: github. You can also contact me via my website. Pull requests are welcome!

Usage

library(logitr)

#we will use the iris data set
X = dplyr::select(iris, -c(Species, Sepal.Length))
y = dplyr::select(iris, Sepal.Length)
y_class = ifelse(dplyr::select(iris, Species) == "setosa", 1, 0)

#ridge regression (use CV for optimal lambda)
linearr(X, y, penalty = "ridge")

#ridge logistic regression (IRLS) (use CV for optimal lambda)
logisticr(X, y_class, penalty = "ridge")

#ridge logistic regression (MM)
logisticr(X, y_class, lam = 0.1, penalty = "ridge", method = "MM")

#bridge logistic regression (MM)
fit = logisticr(X, y_class, lam = 0.1, alpha = 1.2, penalty = "bridge")
fit

#predict using bridge logistic regression estimates
predict(fit, X[1:3,], y_class[1:3])