robCompositions (version 2.6.0)

pls_pb: Function calculating a set of (D-1) principal balances based on PLS.

Description

Function calculating a set of (D-1) principal balances based on PLS.

Usage

pls_pb(Xcoda, ycoda, version = "cov")

Value

A list with the following components:

bal

A matrix of (D-1) principal balances.

cov

Covariance of each balance with the response variable.

Arguments

Xcoda

a matrix of raw compositional data with "n" rows and "D" columns/components

ycoda

a response variable; can be continuous (PLS regression) or binary (PLS-DA)

version

a parameter determining whether the balances are ordered according to max. covariance (default) or max. correlation

Author

Viktorie Nesrstová

Details

The function creates a set of (D-1) principal balances based on PLS. The procedure builds on the method building principal balances based on PCA, introduced in Martin-Fernandez et al. (2018) For detailed information regarding PLS principal balances, see Nesrstová et al. (2023).

References

J. A. Martín-Fernández, V. Pawlowsky-Glahn, J. J. Egozcue, and R. Tolosona-Delgado. Advances in principal balances for compositional data. Mathematical Geosciences, 50(3):273–298, 2018. Available at: https://link.springer.com/article/10.1007/s11004-017-9712-z DOI: tools:::Rd_expr_doi("10.1007/s11004-017-9712-z")

Nesrstová, V, Wilms, I, Palarea-Albaladejo, J, et al. Principal balances of compositional data for regression and classification using partial least squares. Journal of Chemometrics. 2023; 37(12):e3518.. Available at: https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/full/10.1002/cem.3518 DOI: tools:::Rd_expr_doi("10.1002/cem.3518")

Examples

if (FALSE) {
  if (requireNamespace("MASS", quietly = TRUE)) {

# 1. Generate sample data ------------------------------------------------------
n <- 100              # observations
D <- 15               # parts/variables
Sig <- diag(D-1)      # positive-definite symmetric matrix -> covariance matrix
mu <- c(rep(0, D-1))  # means of variables

set.seed(123)
# ilr coordinates
Z <- MASS::mvrnorm(n,mu,Sigma = Sig)

# Z -> CoDa X
V <- compositions::ilrBase(D = D)  # ilrBase() in library(compositions)
X <- as.matrix(as.data.frame(acomp(exp(Z%*%t(V)))))

# Response y:
beta <- runif(D-1,0.1,1)
eps <- rnorm(n)
y <- Z%*%beta+eps

# 2. Calculate PLS PBs

PLS_balances <- fBalChip_PLS(X,y,version = "cov")     # version = "cov" -> max. covariance
balances <- PLS_balances$bal
  }
}