LOG to summarize steps in Discrimination among the 
Bank Notes using PC projections
==================================================

> BankS = var(SwissBNot)
  PCbank = (data.matrix(SwissBNot)- outer(rep(1,200),
             apply(SwissBNot,2,mean),"*")) %*% eigen(BankS)$vec[,1:2]
  > apply(PCbank,2,summary)
              [,1]       [,2]
Min.    -2.724e+00 -3.195e+00
1st Qu. -1.652e+00 -6.111e-01
Median  -2.468e-01 -9.049e-03
Mean     1.108e-15 -3.758e-15
3rd Qu.  1.507e+00  5.804e-01
Max.     3.073e+00  3.164e+00

 > plot(PCbank[,1],PCbank[,2], xlab="PC1 coord", ylab="PC2 coord",
      main="Scatterplot of PC1&2 for BankNotes, 2 Groups", type="n")    
   points(PCbank[1:100,1],PCbank[1:100,2], pch=5)
   points(PCbank[101:200,1],PCbank[101:200,2], pch=18)
### So there is virtually perfect discrimination by this method 
###   using PC1 alone; PC2 is not helpful except perhaps in 
###   helping us to find an oblique line to discriminate perfectly.
  legend(locator(), legend=c("Genuine","Forged"), pch=c(5,18))