hmlasso: Fit a model using a design matrix

In hmlasso: Lasso with High Missing Rate

Description

Fit a model using a design matrix

Usage

hmlasso(X, y, family = "gaussian", impl = "cpp",
  lambda.min.ratio = 0.01, nlambda = 100, lambda = NULL,
  min_eig_th = 1e-06, use = "pairwise.complete.obs",
  positify = "diag", weight_power = 1, eig_tol = 1e-08,
  eig_maxitr = 1e+08, mu = 1, verbose = FALSE, ...)

Arguments

X	matrix of explanatory variables
y	vector of objective variable
family	family of regression: "gaussian" (default) or "binomial"
impl	implementation language of optimization (only "cpp" is supported)
lambda.min.ratio	ratio of max lambda and min lambda (ignored if lambda is specified)
nlambda	the number of lambda (ignored if lambda is specified)
lambda	lambda sequence
min_eig_th	threshold of the minimum eigenvalue in the PSD matrix problem.
use	method to calculate correlation matrix from missing data (default "pairwise.complete.obs")
positify	method for solving PSD matrix
weight_power	weighting power (default 0 meaning no-weighting)
eig_tol	tol parameter in eigs_sym function
eig_maxitr	maxitr parameter in eigs_sym
mu	augmented Lagrangian parameter
verbose	whether output verbose warnings and messages (default FALSE)
...	parameters for optimization

Value

lasso model

beta	coefficients
beta_standard	standardized coefficients
a0	intercepts
lambda	regularization parameters
family	family

Examples

X_incompl <- as.matrix(iris[, 1:3])
X_incompl[1:5,1] <- NA
X_incompl[6:10,2] <- NA
y <- iris[, 4]
fit <- hmlasso(X_incompl, y, nlambda=50, lambda.min.ratio=1e-2)
plot(fit)

hmlasso documentation built on Aug. 3, 2019, 9:03 a.m.