mixlasso: Structured penalized regression
In zhizuio/mixlasso: Mix lasso models

mixlasso

R Documentation

Structured penalized regression

Description

Function producing results of the structured penalized regression

Usage

mixlasso(
  x,
  y,
  z = NULL,
  x_test = NULL,
  y_test = NULL,
  z_test = NULL,
  p = NA,
  foldid = NULL,
  num.nonpen = 0,
  method = "IPF-lasso",
  family = "gaussian",
  nfolds = 5,
  y.mis = NULL,
  y.mis_test = NULL,
  x.mis = NULL,
  x.mis_test = NULL,
  tree.parm = NULL,
  cv.measure = "mse",
  type.measure = "deviance",
  type.min = "lambda.min",
  standardize.response = FALSE,
  lambda = NULL,
  bounds = NULL,
  bound.scale = NA,
  strata.surv = NULL,
  search.path = FALSE,
  EI.eps = 0.01,
  fminlower = 0,
  intercept = TRUE,
  threshold = 0,
  tol = 1e-06,
  mu = 0.01,
  NoVar = 50,
  N = NULL,
  min.iter = 20,
  seed = 1234,
  parallel = FALSE,
  verbose = TRUE,
  t.idx = NULL,
  t.idx_test = NULL,
  t.glasso = FALSE,
  alpha = 1,
  gamma = 0,
  maxiter = 10000,
  cov.proxy = "FL",
  predict.re = FALSE,
  ...
)

Arguments

`x, y`	`x` is the input design matrix; `y` is the input response matrix
`x_test, y_test`	`x` is the input validated design matrix; `y` is the input validated response matrix
`p`	the number of predictors from different data source.
`foldid`	an vector of values for the cross-validation.
`num.nonpen`	number of predictors forced to be estimated (i.e., nonpenalization).
`method`	specify the the method to optimize its penalty parameters. The penalty parameters of `elastic-net`, `IPF-lasso`, `sIPF-elastic-net`, `IPF-elastic-net`, `IPF-tree-lasso` and `clogitLasso` are optimzed by the EPSGO algorithm. The penalty parameter of `lasso` and `tree-lasso` is optimzed by cross-validation. The default method is `IPF-lasso`
`lambda`	optional user-supplied `lambda` sequence; default is NULL, and `espsgo` chooses its own sequence except the tree-lasso methods.
`bounds`	bounds for the interval-searching parameters
`strata.surv`	stratification variable for the Cox survival model.
`search.path`	save the visited points, default is `FALSE`.
`EI.eps`	he convergence threshold for the expected improvement between fmin and the updated point
`fminlower`	minimal value for the function Q.func, default is 0.
`threshold`	threshold for estimated coefficients of the tree-lasso methods.
`N`	define the number of start points depending on the dimensionality of the parameter space.
`min.iter`	the minimus iterations after the initial `N` iterations.
`seed`	random seed.
`parallel`	If `TRUE`, use parallel foreach to fit each fold except parallelizing each lambda for the tree-lasso methods. If `c(TRUE,TRUE)`, use parallel foreach to fit each fold and each lambda.
`verbose`	print the middle search information, default is `TRUE`. ##param lib.loc a character vector describing the location of R library trees to search through, or NULL by default.

Details

mixlasso

Value

An object of list "mixlasso" is returned:

`cvm`	the mean cross-validated error
`cvm_cv`	the mean cross-validated error if providing external dataset "`x_test`" and "`y_test`".
`alpha`	optimized `alpha`
`lambda`	optimized `lambda`
`pred`	the prediction of the responses
`ipf`	optimzed penalty factors
`Beta`	estimate of the coefficients
`cv`	number of nonzero coefficients