yacca (version 1.4-2)

F.test.cca: F Test for Canonical Correlations Using Rao's Approximation

Description

Tests a series of canonical correlations (sequentially) against the null hypothesis that the tested coefficient and all succeeding coefficients are zero.

Usage

F.test.cca(x, …)
# S3 method for F.test.cca
print(x, …)

Arguments

x

a cca object.

…

additional arguments.

Value

An object of class F.test.cca, whose elements are as follows:

corr

Canonical correlations.

statistic

Squared canonical correlations (shared variance across canonical variates).

parameter

Coefficients for the x variables on each canonical variate.

p.value

Coefficients for the y variables on each canonical variate.

method

Canonical variate scores for the x variables.

data.name

Canonical variate scores for the y variables.

Details

Several related tests have been proposed for the evaluation of canonical correlations (including Bartlett's Chi-squared test, which is computed by default within cca). This function employs Rao's statistic (related to Wilks' Lambda) as the basis for an F test of each coefficient (and all others in ascending sequence) against the hypothesis that the associated population correlations are zero.

References

Mardia, K. V.; Kent, J. T.; and Bibby, J. M. 1979. Multivariate Analysis. London: Academic Press.

See Also

cca

Examples

# NOT RUN {
#Example: perceived personal attributes versus professional performance
#for US Judges
data(USJudgeRatings)
personal <- USJudgeRatings[,c("INTG","DMNR","DILG","FAMI","PHYS")]
performance <- USJudgeRatings[,c("CFMG","DECI","PREP","ORAL","WRIT")]
cca.fit <- cca(personal, performance)

#Test the canonical correlations (see also summary(cca.fit))
F.test.cca(cca.fit)
# }