negbin_helpers: Negative Binomial Regression Helpers for lgspline

In lgspline: Lagrangian Multiplier Smoothing Splines for Smooth Function Estimation

Negative Binomial Regression Helpers for lgspline

Description

Functions for fitting negative binomial (NB2) regression models within the lgspline framework. Analogous to the Weibull AFT and Cox PH helpers, these provide the log-likelihood, score, information, and all interface functions needed by lgspline's unconstrained fitting, penalty tuning, and inference machinery.

Details

The parameterization follows NB2: Y \sim \mathrm{NB}(\mu, \theta) with \mathrm{Var}(Y) = \mu + \mu^2/\theta, where \theta > 0 is the shape (size) parameter. The canonical link is log: \eta = \log\mu. The dispersion stored in lgspline$sigmasq_tilde is \theta itself, not 1/\theta.

lgspline documentation built on May 8, 2026, 5:07 p.m.