Association Plot for Contingency Tables

Plots an association plot (Cohen-Friendly plot) for a two-dimensional contingency table. Cells are drawn with widths proportional to the square root of expected frequencies and heights proportional to Pearson residuals. Color encodes both the direction and strength of the association using a diverging palette.

Usage

plotAssoc(
  x,
  main = NULL,
  xlab = TRUE,
  ylab = TRUE,
  space = 0.3,
  reorder = TRUE,
  col = pal("red-white-blue-3", n = 100L),
  border = NA,
  labels = FALSE,
  stamp = .useTheme,
  ...
)

Arguments

x: a two-dimensional contingency table (table or matrix).
main: main title. NULL (default) draws none, unless the theme sets autoTitle = TRUE (see theme); the title is then derived from the expression passed as x (e.g. "tab"). "", NA or FALSE always suppress it and compact the top margin; any other string is used as given.
xlab: character, x-axis label. Defaults to the first dimension name.
ylab: character, y-axis label. Defaults to the second dimension name.
space: numeric, fraction of average cell width/height used as gap between cells. Default 0.3.
reorder: logical. If TRUE (default), rows and columns are reordered by the strength of the strongest association (max(|residual|)) in descending order.
col: character vector of colors for the diverging palette. Default uses pal("red-white-blue-3", n = 100) from the DescToolsX theme. Negative residuals map to the first color, zero to the middle, positive to the last.
border: the color of the border
labels: logical or character. If TRUE, Pearson residuals are printed inside each cell. If FALSE (default), no labels are shown. A character format string (e.g. "\%.1f") can also be passed for custom formatting.
stamp: controls the corner stamp. .useTheme (default) resolves to getTheme()$stamp. TRUE/FALSE/ NULL, or an explicit string, as for .withGraphicsState() (internal).
...: further arguments passed to graphics::rect().

Value

A list with the components x and y, returned invisibly: the horizontal centres of the bars, one per column of the table, and the baselines of the residuals, one per row. Both are named by the levels and given in the order in which they are drawn.

Details

The plot is based on the association plot described in Cohen (1980) and Friendly (1992). Each cell $(i,j)$ is represented by a rectangle:

width proportional to $\sqrt{e_{ij}}$ (square root of expected frequency)
height proportional to the Pearson residual $d_{ij} = (f_{ij} - e_{ij}) / \sqrt{e_{ij}}$

A horizontal baseline at zero represents independence. Cells above the baseline indicate positive association, cells below negative association.

Color encodes both direction and magnitude: the diverging palette runs from the negative color (strong negative residual) through white (no association) to the positive color (strong positive residual).

References

Cohen, A. (1980). On the graphical display of the significant components of a two-way contingency table. Communications in Statistics — Theory and Methods, 9, 1025–1041.

Friendly, M. (1992). Graphical methods for categorical data. SAS User Group International Conference Proceedings, 17, 190–200.

Examples

tab <- table(bedrock::Pizza$driver, bedrock::Pizza$area)

# default
plotAssoc(tab)


# custom palette
plotAssoc(tab, col = pal("red-white-green", n = 100))


# with residual labels
plotAssoc(tab, labels = TRUE)


# no reordering
plotAssoc(tab, reorder = FALSE)


 
plotAssoc(tab,
          main = "Association Driver ~ Area",
          cutoff = 1,
          xlab="Driver", ylab="Area")


cols <- pal()[c(12, 8)]
plotAssoc(tab,
          main = "Association Driver ~ Area",
          cutoff = 1, 
          col = fade(cols, 0.7), border = cols,
          reorder = TRUE, cex.axis = 0.9, 
          xlab = list(labels = "Driver ", 
                    col = "#5B2A45", cex = 1.1), 
          ylab = NA, labels = TRUE)