| reconst {ade4} | R Documentation |
Generic Function for the reconstitution of data from a principal component analysis or a correspondence analysis
reconst (dudi, ...) ## S3 method for class 'pca' reconst(dudi, nf = 1, ...) ## S3 method for class 'coa' reconst(dudi, nf = 1, ...)
dudi |
an object of class |
nf |
an integer indicating the number of kept axes for the reconstitution |
... |
further arguments passed to or from other methods |
returns a data frame containing the reconstituted data
Daniel Chessel
Anne B Dufour anne-beatrice.dufour@univ-lyon1.fr
Gabriel, K.R. (1978) Least-squares approximation of matrices by additive and multiplicative models. Journal of the Royal Statistical Society, B , 40, 186–196.
data(rhone)
dd1 <- dudi.pca(rhone$tab, nf = 2, scann = FALSE)
rh1 <- reconst(dd1, 1)
rh2 <- reconst(dd1, 2)
par(mfrow = c(4,4))
par(mar = c(2.6,2.6,1.1,1.1))
for (i in 1:15) {
plot(rhone$date, rhone$tab[,i])
lines(rhone$date, rh1[,i], lty = 2)
lines(rhone$date, rh2[,i], lty = 1)
ade4:::scatterutil.sub(names(rhone$tab)[i], 2, "topright")}
data(chats)
chatsw <- data.frame(t(chats))
chatscoa <- dudi.coa(chatsw, scann = FALSE)
model0 <- reconst(chatscoa, 0)
round(model0,3)
round(chisq.test(chatsw)$expected,3)
chisq.test(chatsw)$statistic
sum(((chatsw-model0)^2)/model0)
effectif <- sum(chatsw)
sum(chatscoa$eig)*effectif
model1 <- reconst(chatscoa, 1)
round(model1, 3)
sum(((chatsw-model1)^2)/model0)
sum(chatscoa$eig[-1])*effectif