ChemoSpec (version 3.0-1)

mclust3dSpectra: mclust Analysis of a Spectra Object in 3D

Description

This function conducts an mclust analysis of the PCA results of a "Spectra" object and displays the results in 3D. Classical or robust confidence ellipses can be added if desired. Improperly classified data points can be marked. rgl graphics are employed.

Usage

mclust3dSpectra(spectra, pca, pcs = c(1:3),
	ellipse = TRUE, rob = FALSE, cl = 0.95, frac.pts.used = 0.8,
	truth = NULL, title = "no title provided",
	t.pos = NULL, lab.opts = FALSE, use.sym = FALSE, ...)

Arguments

spectra

An object of S3 class "Spectra".

pca

An object of class prcomp.

pcs

An integer vector describing which PCs to use.

ellipse

Logical indicating if confidence ellipses should be drawn.

rob

Logical; if ellipse = TRUE, indicates that robust confidence ellipses should be drawn. If FALSE, classical confidence ellipses are drawn.

cl

A number indicating the confidence interval for the ellipse.

frac.pts.used

If ellipse = TRUE and rob = TRUE, a number indicating the fraction of the data points to be considered "good" and thus used to compute the robust confidence ellipse.

truth

A character vector indicating the known group membership for reach row of the PC scores. Generally this would be spectra$groups.

title

A character string for the plot title.

t.pos

A character selection from LETTERS[1:8] ( = A through H) indicating the desired location for the title.

lab.opts

A logical indicating whether or not to display the locations where the title and legend can be placed. These locations are the corners of a cube surrounding the data.

use.sym

Logical; if true, the color scheme is changed to black and symbols are used for plotting.

...

Other parameters to be passed downstream.

Value

• The mclust model is returned invisibly, and a plot is produced.

Details

If you intend to make a hard copy of your plot, use lab.opts = TRUE until you have found a good view of your data. Then note corners of the cube where the title won't interfere with viewing the data, and use this for t.pos, and add title. Adjust as necessary, then turn off label display using lab.opts = FALSE. Back at the console, use > rgl.snapshot("file_name.png") to create the hardcopy.

Note that the confidence ellipses computed here are generated independently of the Mclust results - they do not correspond to the ellipses seen in 2D plots from Mclust.

References

https://github.com/bryanhanson/ChemoSpec

Examples

data(CuticleIR)
class <- classPCA(CuticleIR, choice = "noscale")
mclust3dSpectra(CuticleIR, class, title = "mclust3dSpectra demo",
	lab.opts = FALSE, t.pos = "A")