Spectral Majorization for Statistical Estimation

Fits generalized linear, asymmetric and robust regression models using spectral majorization and iterative quadratic subproblem solvers. A shared C++ computational core supports dense and sparse designs, automatic intercepts, optional Wald inference and detailed iteration diagnostics. Native code is compiled automatically during package installation.

SpectralMM for R

Fit generalized linear models and robust or asymmetric regression with the shared SpectralMM C++ core. Requires R 4.0 or later and a C++17 compiler for source installation.

install.packages("remotes")
remotes::install_github("Xunjian-Li/SpectralMM", subdir="R")
library(SpectralMM)
model <- spectralmm_glm(mpg ~ wt + hp, data=mtcars, family=gaussian())
summary(model)
predict(model, newdata=mtcars[1:5, ])

For a local checkout, use remotes::install_local("R") from the repository root, or install Rcpp and RcppEigen and run R CMD INSTALL R in a terminal. Edit R code in R/ and help files in man/. Shared C++ files in src/native/ are synchronized from the repository's root cpp/src/native/ directory.

See help(package="SpectralMM") and the full documentation.