Multivariate and cluster analysis
Several standard multivariate methods are provided by GenStat directives. These include
methods that analyse data in the form of units-by-variates, and methods that use a similarity or
distance matrix.
The following directives carry out standard multivariate analyses:
canonical variates analysis
factor analysis
principal components analysis
principal coordinates analysis
Procrustes rotation
non-metric multidimensional scaling
Separate directives are available to process results from multivariate analyses:
rotates factor loadings from a
PCP,
CVA or FCA
adds points for new objects to a
PCO
relates principal coordinates to original data variables
The following directives are used for hierarchical or non-hierarchical cluster analysis:
forms a similarity matrix or a between-group similarity matrix
from a units-by-variates data matrix
forms a reduced similarity matrix (by groups)
hierarchical cluster analysis from a similarity matrix
non-hierarchical clustering from a data matrix
Separate directives that process the results from hierarchical cluster analyses are:
displays results associated with hierarchical clustering
lists a data matrix in abbreviated form
summarizes data variates by clusters
Other multivariate techniques are provided by procedures in the Library:
allows exploratory analysis of genotype × environment
interactions
constructs a classification tree
displays a classification tree
identifies specimens using a classification tree
forms values for nodes of a classification tree
produces a biplot from a set of variates
constructs an identification key
displays an identification key
identifies specimens using a key
identifies an unknown specimen from a defined set of
objects
does canonical correlation analysis
performs canonical correspondence analysis
plots correlation or distance biplots after
CCA or
RDA
plots ordination biplots or triplots after
CCA or
RDA
clusters rows and columns of a two-way interaction
table
obtains a starting classification for non-hierarchical
clustering
finds the points of a single or a full peel of convex-hulls
does correspondence analysis, or reciprocal averaging
(synonym CORRESP)
plots the mean and unit scores from a canonical variates
analysis
calculates scores for individual units in canonical variates
analysis
plots a biplot from an analysis by
PCP,
CVA or PCO
draws dendrograms with control over structure and style
performs discriminant analysis
gives a high resolution plot of an ordination with minumum spanning
tree
displays multivariate data using parallel coordinates
calculates stability coefficients for genotype-by-environment
data
plots displays to assess genotype + genotype-by-environment
variation
performs a generalized Procrustes analysis
prints a scree diagram and/or a difference table of latent
roots
performs multivariate analysis of variance and covariance
assesses the association between similarity matrices
estimates missing values for units in a multivariate data
set
performs tests of univariate and/or multivariate normality
performs a multiple Procrustes analysis
fits a partial least squares regression model
performs redundancy analysis
produces ridge regression and principal component regression
analyses
fits a linear functional relationship model
performs multivariate linear regression with accumulated
testing of terms
forms robust estimates of sum-of-squares-and-products
matrices
produces statistics and graphs for checking sensory panel
performance
provides an analysis of skew-symmetry for an asymmetric
matrix