Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!
Get Started for Free

loocv.discrimin

Leave-one-out cross-validation for a discrimin analysis


Description

Leave-one-out cross-validation to test the existence of groups in a discrimin analysis.

Usage

## S3 method for class 'discrimin'
loocv(x, progress = FALSE, ...)

Arguments

x

the discrimin analysis on which cross-validation should be done

progress

logical to display a progress bar during computations (see the progress package)

...

further arguments passed to or from other methods

Details

This function returns a list containing the cross-validated coordinates of the rows. The analysis on which the discrimin was computed is redone after removing each row of the data table, one at a time. A discrimin analysis is done on this new analysis and the coordinates of the missing row are computed by projection as supplementary element in the new discrimin analysis. This can be useful to check that the groups evidenced by the discrimin analysis are supported.

Value

A list with the cross-validated row coordinates XValCoord, the Predicted Residual Error Sum (PRESS, for each row and PRESSTot, its sum for each discrimin axis), the Root Mean Square Error (RMSE) and the IQR-standardized RMSE (RMSEIQR) for each discrimin axis.

Author(s)

Jean Thioulouse

See Also

Examples

## Not run: 
# Data = skulls
data(skulls)
pcaskul <- dudi.pca(skulls, scan = FALSE)
facskul <- gl(5,30)
diskul <- discrimin(pcaskul, facskul, scan = FALSE)
xdiskul <- loocv(diskul, progress = TRUE)
pst1 <- paste0("Skulls discrimin randtest: p=", round(randtest(diskul)$pvalue, 4))
pst2 <- paste0("Skulls cross-validation: Ax1= ", round(xdiskul$RMSEIQR[1],2),
" Ax2= ", round(xdiskul$RMSEIQR[2],2))
if (adegraphicsLoaded()) {
	sc1 <- s.class(diskul$li, facskul, col = TRUE, psub.text = pst1, ellipseSize=0,
	chullSize=1, plot = FALSE)
	sc2 <- s.class(xdiskul$XValCoord, facskul, col = TRUE, psub.text = pst2,
	ellipseSize=0, chullSize=1, plot = FALSE)
	ADEgS(list(sc1, sc2), layout=c(2,2))
} else {
	par(mfrow=c(2,2))
	s.class(diskul$li, facskul, sub = pst1)
	s.class(xdiskul$XValCoord, facskul, sub = pst2)
}
data(chazeb)
pcacz <- dudi.pca(chazeb$tab, scan = FALSE)
discz <- discrimin(pcacz, chazeb$cla, scan = FALSE)
xdiscz <- loocv(discz, progress = TRUE)
pst1 <- paste0("Chazeb discrimin randtest: p=", round(randtest(discz)$pvalue, 4))
pst2 <- paste0("Chazeb cross-validation: Axis 1= ", round(xdiscz$RMSEIQR[1],2))
if (adegraphicsLoaded()) {
	tabi <- cbind(discz$li, pcacz$tab)
	gr1 <- s.class(tabi, xax=1, yax=2:7, chazeb$cla, col = TRUE, plot = FALSE)
	for (i in 1:6) gr1[[i]] <- update(gr1[[i]], psub.text = names(tabi)[i+1],
	plot = FALSE)
	pos1 <- gr1@positions
	pos1[,1] <- c(0, .3333, .6667, 0, .3333, .6667)
	pos1[,2] <- c(.6667, .6667, .6667, .3333, .3333, .3333)
	pos1[,3] <- c(.3333, .6667, 1, .3333, .6667, 1)
	pos1[,4] <- c(1, 1, 1, .6667, .6667, .6667)
	gr1@positions <- pos1
	sc1 <- s1d.gauss(discz$li, chazeb$cla, col = TRUE, psub.text = pst1,
	plot = FALSE)
	sc2 <- s1d.gauss(xdiscz$XValCoord, chazeb$cla, col = TRUE, psub.text = pst2,
	plot = FALSE)
	ADEgS(list(gr1[[1]], gr1[[2]], gr1[[3]], gr1[[4]], gr1[[5]], gr1[[6]], sc1, sc2))
} else {
	plot(discz)
	sco.gauss(discz$li[,1], as.data.frame(chazeb$cla), sub = pst1,
	legen = FALSE)
	sco.gauss(xdiscz$XValCoord[,1], as.data.frame(chazeb$cla), sub = pst2,
	legen = FALSE)
}

## End(Not run)

ade4

Analysis of Ecological Data: Exploratory and Euclidean Methods in Environmental Sciences

v1.7-16
GPL (>= 2)
Authors
Stéphane Dray <stephane.dray@univ-lyon1.fr>, Anne-Béatrice Dufour <anne-beatrice.dufour@univ-lyon1.fr>, and Jean Thioulouse <jean.thioulouse@univ-lyon1.fr>, with contributions from Thibaut Jombart, Sandrine Pavoine, Jean R. Lobry, Sébastien Ollier, Daniel Borcard, Pierre Legendre, Stéphanie Bougeard and Aurélie Siberchicot. Based on earlier work by Daniel Chessel.
Initial release

We don't support your browser anymore

Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.