gamclass (version 0.62.5)

bssBYcut: Between group SS for y, for all possible splits on values of x

Description

Each point of separation between successve values of x is used in turn to create two groups of observations. The between group sum of squares for y is calculated for each such split.

Usage

bssBYcut(x, y, data)

Value

Data frame with columns:

xOrd

Cut points for splits.

comp2

Between groups sum of squares

Arguments

x

Variable (numeric) used to define splits. Observations with x values less than the cut point go into the first group, while those with values >= the cut point go into the second group.

y

Variable for which BSS values are to be calculated.

data

Data frame with columns x and y.

Author

J H Maindonald

Examples

xy <- bssBYcut(weight, height, women)
with(xy, xy[which.max(bss), ])

## The function is currently defined as
function (x, y, data)
{
    xnam <- deparse(substitute(x))
    ynam <- deparse(substitute(y))
    xv <- data[, xnam]
    yv <- data[, ynam]
    sumss <- function(x, y, cut) {
        av <- mean(y)
        left <- x < cut
        sum(left) * (mean(y[left]) - av)^2 + sum(!left) * (mean(y[!left]) -
            av)^2
    }
    xOrd <- unique(sort(xv))[-1]
    bss <- numeric(length(xOrd))
    for (i in 1:length(xOrd)) {
        bss[i] <- sumss(xv, yv, xOrd[i])
    }
    list(xOrd = xOrd, bss = bss)
  }