# load swiss bank notes
x<-read.table("bank2.dat")
spr <-princomp(x)
U<-spr$loadings
L<-(spr$sdev)^2
Z <-spr$scores
# scatterplot of input x
pairs(x,pch=c(rep(1,100),rep(3,100)),col=c(rep("blue",100),rep("red",100)))
# scatterplot of principal components
pairs(Z,pch=c(rep(1,100),rep(3,100)),col=c(rep("blue",100),rep("red",100)))
# variances of each principal component
par(mfrow=c(1,2))
plot(L,type="b",xlab="component",ylab="lambda",main="Scree plot")
plot(cumsum(L)/sum(L)*100,ylim=c(0,100),type="b",xlab="component",ylab="cumulative propotion (%)",main="Cum. Scree plot")
# biplot
par(mfrow=c(1,1))
biplot(spr, choices = 1:2, scale = 1, pc.biplot = FALSE)

par(mfrow=c(1,2))
barplot(spr$loadings[,1],main="PC1 loadings")
barplot(spr$loadings[,2],main="PC2 loadings")
 

# preprocessed golub data
#source("http://bioconductor.org/biocLite.R")
#biocLite("multtest")
data(golub, package = "multtest") 
dim(golub)
# golub.gnames # names of genes
# golub.cl     # indicater for ALL (AML==0, ALL==1)

# gpr <- princomp(t(golub))
gpr <- prcomp(t(golub))
U <- gpr$rotation
L <- (gpr$sdev)^2
Z <- gpr$x

# scatterplot of principal components
pairs(Z[,1:4],pch=c(rep(1,27),rep(3,11)),col=c(rep("blue",27),rep("red",11)))
pairs(Z[,1:8],pch=c(rep(1,27),rep(3,11)),col=c(rep("blue",27),rep("red",11)))
# variances of each principal component
par(mfrow=c(1,2))
plot(L,type="b",xlab="component",ylab="lambda",main="Scree plot")
plot(cumsum(L)/sum(L)*100,ylim=c(0,100),type="b",xlab="component",ylab="cumulative propotion (%)",main="Cum. Scree plot")
 # biplot does not help
# biplot
par(mfrow=c(1,1))
biplot(gpr, choices = 1:2, scale = 1, pc.biplot = FALSE)