Abstract:
|
Ordinal regression models arise in contexts where the response variable belongs to one of several ordered categories (such as 1="poor", 2="fair", 3="good", 4="excellent"). Ordinal logistic regression (also known as proportional odds or ordered logit regression) is widely used in applications where the use of regularization and variable selection could be beneficial. However, ordinal logistic regression is not included in many popular software packages for regularized regression. We propose a coordinate descent algorithm to fit ordinal logistic regression with elastic net penalty. We also introduce the R package ordinalNet, which implements the algorithm.
|