R/dgower.R

In ICGE: Estimation of Number of Clusters and Identification of Atypical Units

dgower <- function(x, type=list()){
############ Gower distance for mixed data ###############
# Input:
# x: data matrix. Missing values are indicated as "NA"
# type: list with components,
#      cuant =  position of quantitative variables
#      bin = position of  binary variables (asymmetric)
#      nom = position of nominal variables
# Output:
# d: distance matrix
###########################################################

    dx <- dim(x)
    n <- dx[1]
    p <- dx[2]
    ntype <- names(type)
    if (length(type)) {
        if ((!is.list(type)) || (is.null(ntype)) || any(ntype == 
                                                        "")) {
            stop("invalid ", sQuote("type"), "; must be named list")
        }
        for (nt in ntype) {
            cvec <- type[[nt]]
            if (is.numeric(cvec)) {
                if (!all(1 <= cvec & cvec <= p)) {
                    stop("type$", nt, " must be in 1:ncol(x)")
                }
            }
            else {
                stop("type$", nt, " must  contain numbers")
            }
        }
        vc <- type$cuant
        vb <- type$bin
        vn <- type$nom
        v <- c(vc, vb, vn)
        if (!all(sort(v) == 1:p)) {
            stop("type must contain speficications for all variables")
        }
    }
    else {
        stop("type must contain speficications for all variables")
    }
    if (any(is.na(x))) {
        missing <- TRUE
    }
    else {
        missing <- FALSE
    }
    if (missing) {
        d <- gowerWITHmissing_v2(x, vc, vb, vn)
    }
    else {
        d <- gowerNOmissing_v2(x, vc, vb, vn)
    }
    d <- as.dist(d)
    return(d)
}