cgpaths: Get a solution path for the CG model

In amcdavid/HurdleNormal: 'Estimation and sampling for multi-dimensional Hurdle models on a Normal density with applications to single-cell co-expression'

Get a solution path for the CG model

Description

Get a solution path for the CG model

Usage

cgpaths(
  y.zif,
  this.model,
  Blocks = Block(this.model),
  nodeId = NA_character_,
  nlambda = 50,
  lambda.min.ratio = if (length(y.zif) < ncol(this.model)) 0.005 else 0.05,
  lambda,
  penaltyFactor = "full",
  control = list(tol = 0.005, maxrounds = 300, debug = 1),
  theta
)

Arguments

y.zif	(zero-inflated) response
this.model	model matrix used for both discrete and continuous linear predictors
Blocks	output from Block giving the grouping/scaling for the penalization.
nodeId	optional labels for the nodes. will be used to stitch to
nlambda	if 'lambda' is not provided, then the number of lambda to interpolate between
lambda.min.ratio	if 'lambda' is not provided, then the left end of the solution path as a function of the lambda0, the lambda for the empty model
lambda	penalty path desired
penaltyFactor	one of 'full', 'diagonal' or 'identity' giving how the penalty should be scaled blockwise
control	optimization control parameters
theta	(optional) initial guess for parameter

Value

matrix of parameters, one row per lambda

amcdavid/HurdleNormal documentation built on May 14, 2022, 11:12 p.m.