R/fishers-z.R

In davebraze/FDB1: A brew of utility and convenience functions

##' @title Fisher's Z test for comparing two correllation coefficients.
##'
##' @details
##' Test whether two correlation coefficients are statistically different using the difference
##' between Fisher's Z transforms.
##'
##' Method cribbed from:
##' \enumerate{
##'     \item \url{http://www.fon.hum.uva.nl/Service/Statistics/Two_Correlations.html}
##'     \item \url{http://ftp.sas.com/techsup/download/stat/compcorr.html}
##' }
##' Implemented in R by me.
##'
##' @param n1 n for correlation 1.
##' @param r1 r for correlation 1.
##' @param n2 n for correlation 2.
##' @param r2 r for correlation 2.
##' @return list containing diff(Z1, Z2) and pval for difference.
##' @author Dave Braze \email{davebraze@@gmail.com}
##' @export
compcorr <- function(n1, r1, n2, r2){
    ## Fisher Z-transforms
    zf1 <- 0.5*log((1 + r1)/(1 - r1))
    zf2 <- 0.5*log((1 + r2)/(1 - r2))
    ## difference
    dz <- (zf1 - zf2)/sqrt(1/(n1 - 3) + (1/(n2 - 3)))
    ## p-value
    pv <- 2*(1 - pnorm(abs(dz)))
    retval <- list(diff=dz, pval=pv)
    retval
}