variancePartition (version 1.2.5)

canCorPairs: canCorPairs

Description

Assess correlation between all pairs of variables in a formula

Usage

canCorPairs(formula, data)

Arguments

formula

standard linear model formula (doesn't support random effects currently, so just change the syntax)

data

data.frame with the data for the variables in the formula

Value

Matrix of correlation values between all pairs of variables.

Details

Canonical Correlation Analysis (CCA) is similar to correlation between two vectors, except that CCA can accommodate matricies as well. For a pair of variables, canCorPairs assesses the degree to which they co-vary and contain the same information. Variables in the formula can be a continuous variable or a discrete variable expanded to a matrix (which is done in the backend of a regression model). For a pair of variables, canCorPairs uses CCA to compute the correlation between these variables and returns the pairwise correlation matrix.

Statistically, let rho be the array of correlation values returned by the standard R function cancor to compute CCA. canCorPairs returns rho / sum(rho) which is the fraction of the maximum possible correlation.

Note that CCA returns correlations values between 0 and 1

Examples

# load library
# library(variancePartition)

# load simulated data:
data(varPartData)

# specify formula
form <- ~ Individual + Tissue + Batch + Age + Height

# Compute Canonical Correlation Analysis (CCA)
# between all pairs of variables
# returns absolute correlation value  
C = canCorPairs( form, info)

# Plot correlation matrix
plotCorrMatrix( C )