var_select: Select predictors

In lbraglia/lbstat: Miscellaneous Stat-Related Utils

Select predictors

Description

Select best predictors using alternatively stepwise AIC, ridge regression, Lasso (with optimal lambda determined according to crossvalidation for the latters) or simplified purposeful selection (aka start from univariate of each covariate specified in formula and select what to put in a final multivariate analysis)

Usage

var_select(
  formula,
  data,
  family = c("gaussian", "binomial", "poisson", "multinomial", "cox", "mgaussian"),
  method = c("lasso", "ridge", "step", "purposeful"),
  cv_nfolds = 10,
  lambdas = NULL,
  seed = 1,
  purposeful_p = 0.2
)

Arguments

formula	the full model specification
data	dataset for search
family	type of variable predicted
method	which method to use ('lasso', 'ridge', 'step', 'purposeful' or abbreviations)
cv_nfolds	crossvalidation subsamples for lasso and ridge
lambdas	lambdas for lasso and ridge
seed	seed for pseudo-random number generation
purposeful_p	p-value in univariate analysis below which a variable is included in multivariate analysis

lbraglia/lbstat documentation built on Sept. 17, 2023, 2:30 p.m.