adapt4pv (version 0.2-3)

lasso_perm: fit a lasso regression and use standard permutation of the outcome for variable selection

Description

Performed K lasso logistic regression with K different permuted version of the outcome. For earch of the lasso regression, the \(\lambda_max\)(i.e. the smaller \(\lambda\) such as all penalized regression coefficients are shrunk to zero) is obtained. The median value of these K \(\lambda_max\) is used to for variable selection in the lasso regression with the non-permuted outcome. Depends on the glmnet function from the package glmnet.

Usage

lasso_perm(x, y, K = 20, keep = NULL, betaPos = TRUE, ncore = 1, ...)

Value

An object with S3 class "log.lasso".

beta

Numeric vector of regression coefficients in the lasso In lasso_perm function, the regression coefficients are PENALIZED. Length equal to nvars.

selected_variables

Character vector, names of variable(s) selected with the lasso-perm approach. If betaPos = TRUE, this set is the covariates with a positive regression coefficient in beta. Else this set is the covariates with a non null regression coefficient in beta. Covariates are ordering according to magnitude of their regression coefficients absolute value.

Arguments

x

Input matrix, of dimension nobs x nvars. Each row is an observation vector. Can be in sparse matrix format (inherit from class "sparseMatrix" as in package Matrix).

y

Binary response variable, numeric.

K

Number of permutations of y. Default is 20.

keep

Do some variables of x have to be permuted in the same way as y? Default is NULL, means no. If yes, must be a vector of covariates indices. TEST OPTION

betaPos

Should the covariates selected by the procedure be positively associated with the outcome ? Default is TRUE.

ncore

The number of calcul units used for parallel computing. Default is 1, no parallelization is implemented.

...

Other arguments that can be passed to glmnet from package glmnet other than family.

Author

Emeline Courtois
Maintainer: Emeline Courtois

Details

The selected \(\lambda\) with this approach is defined as the closest \(\lambda\) from the median value of the K \(\lambda_max\) obtained with permutation of the outcome.

References

Sabourin, J. A., Valdar, W., & Nobel, A. B. (2015). "A permutation approach for selecting the penalty parameter in penalized model selection". Biometrics. 71(4), 1185–1194, tools:::Rd_expr_doi("10.1111/biom.12359")

Examples

set.seed(15)
drugs <- matrix(rbinom(100*20, 1, 0.2), nrow = 100, ncol = 20)
colnames(drugs) <- paste0("drugs",1:ncol(drugs))
ae <- rbinom(100, 1, 0.3)
lp <- lasso_perm(x = drugs, y = ae, K = 10)