big.simplelasso: Perform LASSO to select best features for big.matrix objects

In EpidemiumOpenCancer/OpenCancer: Build statistical models to understand cancer causes

Description

big.simplelasso estimates a model using LASSO and returns a sparse structure. Cross-validation can be used to select best covariates combination. The dataframe must be a C++ pointer

Usage

big.simplelasso(X, yvar = "incidence", labelvar = c("cancer", "age",
  "Country_Transco", "year", "area.x", "area.y"), crossvalidation = T,
  nfolds = 10, returnplot = T, ncores = 1)

Arguments

X	A big.matrix object
yvar	Column name for the explained variable
labelvar	Additional variables providing information but that should not be used as explanatory variables
crossvalidation	Should cross-validation be performed? TRUE or FALSE
nfolds	Number of folds for cross-validation. Ignored if crossvalidation = T
ncores	Number of cores to used for computation. If ncores>1, parallel processing is used

Value

A list of three elements.

output$model returns the model.

output$plot returns a plot. If crossvalidation = F, coefficients values when $\lambda$ penalization term evolves is represented. If crossvalidation = T, the RMSE is represented with respect to the number of variables with non-zero weight

output$coeff returns the coefficient values returned by the LASSO (or the coefficients of the RMSE-minimizing model if crossvalidation = T)

EpidemiumOpenCancer/OpenCancer documentation built on May 12, 2019, 7:46 a.m.