R/mgroup.R

In phiala/ecodist: Dissimilarity-Based Functions for Ecological Analysis

# mgroup returns the Mantel correlations for group contrast
# matrices computed from cluster groups across a range of 
# clustering levels. 
# The inputs are an ecological distance matrix
# and a vector or matrix of cluster levels in any order. 
#
# dlu
#

mgroup <- function(edist, groups, nperm = 1000, mrank = FALSE)  
{
    nl <- ncol(groups)
    if(is.null(nl)) {
        groups <- matrix(groups, ncol=1)
        nl <- 1
    }

    if(mrank) {
        edist <- rank(edist)
    }

    outtable <- data.frame(nclust = rep(NA, nl), mantelr = rep(NA, nl), pval = rep(NA, nl))

    for (i in seq_len(nl)) {
        # number of groups at this level:
        thisgroups <- groups[,i]
        if(is.factor(thisgroups)) {
            thisgroups <- as.numeric(thisgroups)
        }
        if(is.character(thisgroups)) {
            thisgroups <- as.numeric(factor(thisgroups))
        }

        cl <- length(unique(groups[,i]))
        # create group contrast:
        gdist <- dist(thisgroups)
        gdist[gdist > 0] <- 1

        # run mantel:
        m <- mantel(edist ~ gdist, nperm=nperm, nboot=0) 
        outtable[["nclust"]][i] <- cl
        outtable[["mantelr"]][i] <- m[1]
        outtable[["pval"]][i] <- m[2]
    }

    outtable
}