fitGlbin_CV: Group Lasso Penalised Logistic Regression Estimation
In antiphon/PenGE: Penalised Gibbs Estimation

Uses the simplified LCD

fitGlbin_CV(Q, dfmax = ncol(Q$X) * 1.1, lambda = NULL,
  lambda.min = 0.001, nlambda = 100, log_linear_lambda = FALSE,
  verb = 0, add_intercept = TRUE, quads = NULL, mc.cores = 3,
  sparse = FALSE, use_glm_in_lmax = TRUE, ...)

`Q`	Model object
`dfmax`	stop if dfmax significant terms reached
`lambda`	penalty vector. if missing, will be computed
`lambda.min`	If lambda is to be computed, min. lambda will be max.lambda * lambda.min
`nlambda`	steps to generate for lambda vector
`log_linear_lambda`	log-linear or linear
`verb`	Verbosity
`add_intercept`	Add overall intercept?
`quads`	The cross validation quadrats
`mc.cores`	mclapply mc.cores
`sparse`	use Sparse Matrix
`use_glm_in_lmax`	use R-s own glm function (non-sparse) to determine max. lambda
`...`	passed on to

Basically just a wrapper for package glbinc for estimating group penalised binomial regression. If 'quads' (list of sub-windows, see 'split_window'-function) are given, fits the CV-models as well. Here the mc.cores -parameter is used.

antiphon/PenGE documentation built on July 31, 2019, 10:01 p.m.