McMiso: Multicore Multivariable Isotonic Regression

Provides functions for isotonic regression and classification when there are multiple independent variables. The functions solve the optimization problem using a projective Bayes approach with recursive sequential update algorithms, and are useful for situations with a relatively large number of covariates. Supports binary outcomes via a Beta-Binomial conjugate model ('miso', 'PBclassifier') and continuous outcomes via a Normal-Inverse-Chi-Squared conjugate model ('misoN'). Parallel computing wrappers ('mcmiso', 'mcPBclassifier', 'mcmisoN') are provided that run the down-up and up-down algorithms simultaneously and return whichever finishes first. The estimation method follows the projective Bayes solution described in Cheung and Diaz (2023) <doi:10.1093/jrsssb/qkad014>.

install.packages("McMiso")

Try the McMiso package in your browser