statforbiology (version 0.9.9)

CVA: Canonical variate analysis for multienvironment and multitrait genotype experiments

Description

This function performs canonical variate analysis as a descriptive visualisation tool. It is close to the 'lda()' function in the MASS package but it is not meant to be used for discriminant analyses.

Usage

CVA(dataset, groups, scale = TRUE, constraint = 3)

Value

TOT

matrix of total variances-covariances

B

matrix of 'between-groups' variances-covariances

W

matrix of 'within-group' variances-covariances

B/W

matrix of W^{-1} B

eigenvalues

vector of eigenvalues

eigenvectors

matrix of eigenvectors

proportion

a vector containing the proportion of total discriminating ability captured by each canonical variate

correlation

vector of canonical correlations

squared.canonical.correlation

vector of squared canonical correlations

coefficients

matrix of canonical coefficients

scores

matrix of canonical scores

centroids

matrix of scores for centroids

total.structure

matrix of total canonical structure

between.structure

matrix of between-groups canonical structure

within.structure

matrix of within-groups canonical structure

class.fun

matrix of classifications functions

class.val

matrix of classification values

within.structure

matrix of within-groups canonical structure

class

vector of predicted classes

Arguments

dataset

dataset is a multidimensional matrix of observations

groups

groups is a vector coding for groupings

scale

whether the data needs to be standardised prior to analysis. Defaults to TRUE

constraint

It is the type of scaling for eigenvectors, so that canonical variates have: 1 = unit within-group standard deviations (most common); 2 = unit total standard deviations; 3 = unit within group norms; 4 = unit total norms. It defaults to 3

Author

Andrea Onofri

Details

More detail can be found in a blog page, at 'https://www.statforbiology.com/2023/stat_multivar_cva/'. Please, note that preliminary data transformations (e.g.: standardisation) are left to the user and must be performed prior to analyses (see example below).

References

https://www.statforbiology.com/2023/stat_multivar_cva/

Examples

dataset <- getAgroData("WheatQuality4years")
dataset$Year <- factor(dataset$Year)
head(dataset)

# Standardise the data
groups <- dataset$Genotype
Z <- apply(dataset[,3:6], 2, scale, center = TRUE, scale = TRUE)
head(Z)

# Performs CVA
cvaobj <- CVA(Z, groups)
cvaobj