se0_ridge: Logistic ridge regression state evolution functions with no...

In brglm2: Bias Reduction in Generalized Linear Models

Logistic ridge regression state evolution functions with no intercept

Description

Logistic ridge regression state evolution functions with no intercept

Usage

se0_ridge(mu, b, sigma, kappa, gamma, lambda, gh = NULL, prox_tol = 1e-10)

Arguments

mu	aggregate bias parameter.
b	parameter b in the state evolution functions.
sigma	square root of the aggregate variance of the MDYPL estimator.
kappa	asymptotic ratio of columns/rows of the design matrix. kappa should be in ⁠(0, 1)⁠.
gamma	the square root of the limit of the variance of the linear predictor.
lambda	the shrinkage parameter of the logistic regression penalty estimator. lambda should be in greater than zero.
gh	A list with the Gauss-Hermite quadrature nodes and weights, as returned from statmod::gauss.quad() with kind = "hermite". Default is NULL, in which case gh is set to statmod::gauss.quad(200, kind = "hermite") is used.
prox_tol	tolerance for the computation of the proximal operator; default is 1e-10.

Details

It is assumed that the ridge penalty to the logistic regression log-likelihood is n * lambda * sum(beta^2) / (2 * length(beta)), where n is the sum of the binomial totals.

brglm2 documentation built on April 29, 2026, 5:06 p.m.