Cross-validation for LASSO with compositional predictors using the alpha-transformation {Compositional}

Cross-validation for LASSO with compositional predictors using the alpha-transformation

Description

Cross-validation for LASSO with compositional predictors using the alpha-transformation.

Usage

alfalasso.tune(y, x, a = seq(-1, 1, by = 0.1), model = "gaussian", lambda = NULL,
type.measure = "mse", nfolds = 10, folds = NULL, stratified = FALSE)

Arguments

y

A numerical vector or a matrix for multinomial logistic regression.

x

A numerical matrix containing the predictor variables, compositional data, where zero values are allowed..

a

A vector with a grid of values of the power transformation, it has to be between -1 and 1. If zero values are present it has to be greater than 0. If \alpha=0 the isometric log-ratio transformation is applied.

model

The type of the regression model, "gaussian", "binomial", "poisson", "multinomial", or "mgaussian".

lambda

This information is copied from the package glmnet. A user supplied lambda sequence. Typical usage is to have the program compute its own lambda sequence based on nlambda and lambda.min.ratio. Supplying a value of lambda overrides this. WARNING: use with care. Avoid supplying a single value for lambda (for predictions after CV use predict() instead). Supply instead a decreasing sequence of lambda values. glmnet relies on its warms starts for speed, and its often faster to fit a whole path than compute a single fit.

type.measure

This information is taken from the package glmnet. The loss function to use for cross-validation. For gaussian models this can be "mse", "deviance" for logistic and poisson regression, "class" applies to binomial and multinomial logistic regression only, and gives misclassification error. "auc" is for two-class logistic regression only, and gives The area under the ROC curve. "mse" or "mae" (mean absolute error) can be used by all models.

nfolds

The number of folds. Set to 10 by default.

folds

If you have the list with the folds supply it here. You can also leave it NULL and it will create folds.

stratified

Do you want the folds to be created in a stratified way? TRUE or FALSE.

Details

The function uses the glmnet package to perform LASSO penalised regression. For more details see the function in that package.

Value

A matrix with two columns and number of rows equal to the number of \alpha values used. Each row contains, the optimal value of the \lambda penalty parameter for the LASSO and the optimal value of the loss function, for each value of \alpha.

Author(s)

Michail Tsagris.

R implementation and documentation: Michail Tsagris [e-mail].

References

Aitchison J. (1986). The statistical analysis of compositional data. Chapman & Hall.

Friedman, J., Hastie, T. and Tibshirani, R. (2010) Regularization Paths for Generalized Linear Models via Coordinate Descent. Journal of Statistical Software, Vol. 33(1), 1–22.

See Also

alfa.lasso, cv.lasso.klcompreg, lasso.compreg, alfa.knn.reg

Examples

y <- iris[, 1]
x <- rdiri(150, runif(20, 2, 5) )
mod <- alfalasso.tune( y, x, a = c(0.2, 0.5, 1) )