RFgroove (version 1.1)

selectFunctional: Grouped variable selection procedure for functional data

Description

Backward variable selection procedure for multivariate functional data which is a set of $p$ functional covariates and $n$ observations of dimension $N$.

Usage

selectFunctional( FDlist, ydata, normalize=TRUE,  dimensionReductionMethod=c("fpca", "wave"),  nbasisInit, verbose=TRUE, ...)

Arguments

FDlist

A p-dimensional list containing the set of functional variables which are matrices of size $n \times N$.

ydata

The outcome data. Must be a factor for classification.

normalize

Should the functions be normalized ?

dimensionReductionMethod

The dimension reduction method, ‘fpca’ for Functional Principal Component Analysis or ‘wave’ for the multiple wavelet thresholding.

nbasisInit

The number of initial spline coefficients.

verbose

Should the details be printed.

...

further arguments passed to or from other methods.

Value

An object of class fRFE which is a list with the following components:

nselected

The number of selected functional variables ;

selection

The selected functional variables ;

selectionIndexes

The indexes of selected functional variables in the input data ‘FDlist’ ;

error

The prediction error computed in each iteration of the backward procedure ;

typeRF

The type of the forests, classification or regression ;

ranking

The final ranking of the functional variables ;

rankingIndexes

The final ranking indexes of the functional variables.

References

Gregorutti, B., Michel, B. and Saint Pierre, P. (2015). Grouped variable importance with random forests and application to multiple functional data analysis, Computational Statistics and Data Analysis 90, 15-35.

See Also

selectGroup,varImpGroup

Examples

  data(toyRegFD)
  varSel <- selectFunctional( toyRegFD$FDlist, toyRegFD$Y, normalize=FALSE, 
                              dimensionReductionMethod="fpca", nbasisInit=16, 
                              verbose=FALSE, ntree=10)
  summary(varSel)
  plot(varSel)