b_update: Update beta estimates using Newton-Raphson algorithm

In theandyb/aeffp: Lasso-penalized Logistic Regression via ADMM

Description

The beta-update step requires optimizing a convex function. This version of the update function uses a Newton-Raphson approach to minimizing the objective function.

Usage

b_update(X, y, u, z, rho, maxiter = 50, toler = 1e-05, b = 0.5,
  alpha = 0.1)

Arguments

X	Covariate matrix (no column for intercept)
y	Vector of observations (coded in -1/1)
u	Current value of u vector (ADMM optimization)
z	Current value of z vector (ADMM optimization)
rho	Tuning parameter for ADMM optimization
maxiter	Maximum number of iterations
toler	Convergence criterion
b	Backtracking tuning parameter
alpha	Backtracking tuning parameter

Value

Vector containing updated estimate of beta vector

theandyb/aeffp documentation built on May 8, 2019, 9:09 a.m.