TPLS_cv {TPLSr}

Constructor method for fitting a cross-validation T-PLS model

Description

Constructor method for fitting a cross-validation T-PLS model

Usage

TPLS_cv(X, Y, CVfold, NComp = 25, W = NULL, nmc = 0)

Arguments

X

Numerical matrix of predictors. Typically single-trial betas where each column is a voxel and row is observation

Y

Variable to predict. Binary 0 and 1 in case of classification, continuous variable in case of regression

CVfold

Cross-validation testing fold information. Can either be a vector or a matrix, the latter being more general. Vector: n-by-1 vector. Each element is a number ranging from 1 ~ numfold to identify which testing fold each observation belongs to Matrix: n-by-numfold matrix. Each column indicates the testing data with 1 and training data as 0. Example: For leave-one-out CV, Vector would be 1:n, Matrix form would be eye(n) Matrix form is more general as it can have same trial be in multiple test folds

NComp

(Optional) Number of PLS components to compute. Default is 25.

W

(Optional) Observation weights. Optional input. By default, all observations have equal weight. Can either be a n-by-1 vector or a n-by-nfold matrix where each column is observation weights in that CV fold

nmc

(Optional) 'no mean centering'. See TPLS for more detail. Turning this on will skip mean centering on all cross validation folds, so they should all be mean-centered already

Value

A TPLS_cv object that contains the following attributes. Most of the time, you won't need to access the attributes.

• NComp: The number of components you specified in the input

• numfold: Total number of cross-validation folds

• CVfold: A matrix of indicators for testing data for each cross validation fold in each column

• cvMdls : A vector of TPLS models, one for each fold.

See vignettes for tutorial