heplots: plot.boxM – R documentation

Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!

plot.boxM

Plot for Box's M test and generalizations

Description

This function creates a simple dot chart showing the contributions (log determinants) of the various groups to Box's M test for equality of covariance matrices.

Because Box's M test is based on a specific function (log determinant) of the covariance matrices in the groups compared to the pooled covariance matrix, this function also also allow plots of other measures based on the eigenvalues of these covariance matrices.

Usage

## S3 method for class 'boxM'
plot(x,
             gplabel = NULL,
             which = c("logDet", "product", "sum", "precision", "max"),
             log = which=="product",
             pch = c(16, 15), 
             cex = c(2, 2.5), 
             col = c("blue", "red"),  
             rev = FALSE,
             xlim,
             conf=0.95, method=1, bias.adj=TRUE, lwd=2,
             ...)

Arguments

`x`	A `"boxM"` object resulting from `boxM`
`gplabel`	character string used to label the group factor.
`which`	Measure to be plotted. The default, `"logDet"`, is the standard plot. Other values are: `"product"`, `"sum"`, `"precision"` and `"max"`
`log`	logical; if `TRUE`, the log of the measure is plotted. The default, `which=="product"`, produces a plot equivalent to the plot of `"logDet"`.
`pch`	a vector of two point symbols to use for the individual groups and the pooled data, respectively
`cex`	character size of point symbols, a vector of length two for groups and pooled data, respectively
`col`	colors for point symbols, a vector of length two for the groups and the pooled data
`rev`	logical; if `TRUE`, the order of the groups is reversed on the vertical axis.
`xlim`	x limits for the plot
`conf`	coverage for approximate confidence intervals, `0 <= conf < 1` ; use `conf=0` to suppress these
`method`	confidence interval method; see `logdetCI`
`bias.adj`	confidence interval bias adjustment; see `logdetCI`
`lwd`	line width for confidence interval
`...`	Arguments passed down to `dotchart`.

Details

Confidence intervals are only available for the default Box M test, using which="logDet".

Author(s)

Michael Friendly

Examples

# Iris data
res <- boxM(iris[, 1:4], iris[, "Species"])
plot(res, gplabel="Species")

# Skulls data
skulls.mod <- lm(cbind(mb, bh, bl, nh) ~ epoch, data=Skulls)
skulls.boxm <- boxM(skulls.mod)
plot(skulls.boxm, gplabel="Epoch")
plot(skulls.boxm, gplabel="Epoch", bias.adj=FALSE)

# other measures
 plot(skulls.boxm, which="product", gplabel="Epoch", xlim=c(10,14))
 plot(skulls.boxm, which="sum", gplabel="Epoch")
 plot(skulls.boxm, which="precision", gplabel="Epoch")
 plot(skulls.boxm, which="max", gplabel="Epoch")

heplots

Visualizing Hypothesis Tests in Multivariate Linear Models

v1.3-8

GPL (>= 2)

Authors

John Fox [aut], Michael Friendly [aut, cre], Georges Monette [ctb], Phil Chalmers [ctb]

Initial release

2021-01-20