This function compares a robust covariance (correlation) estimation (MCD is used) with the classical approach. A plot with the two ellipses will be produced and the correlation coefficients are quoted.
RobCor.plot(x, y, quan = 1/2, alpha = 0.025, colC = 1, colR = 1, ltyC = 2,
ltyR = 1, ...)
correlation of the classical estimation
correlation of the robust estimation
two data vectors where the correlation should be computed
fraction of tolerated outliers (at most 0.5)
quantile of chisquare distribution for outlier cutoff
colour for both ellipses
line type for both ellipses
other graphical parameters
Peter Filzmoser <P.Filzmoser@tuwien.ac.at> http://cstat.tuwien.ac.at/filz/
The covariance matrix is estimated in a robust (MCD) and non robust way and then both ellipses are plotted. The radi is calculated from the singular value decomposition and a breakpoint (specified quantile) for outlier cutoff.
C. Reimann, P. Filzmoser, R.G. Garrett, and R. Dutter: Statistical Data Analysis Explained. Applied Environmental Statistics with R. John Wiley and Sons, Chichester, 2008.
data(chorizon)
Be=chorizon[,"Be"]
Sr=chorizon[,"Sr"]
RobCor.plot(log10(Be),log10(Sr),xlab="Be in C-horizon [mg/kg]",
ylab="Sr in C-horizon [mg/kg]",cex.lab=1.2, pch=3, cex=0.7,
xaxt="n", yaxt="n",colC=1,colR=1,ltyC=2,ltyR=1)
Run the code above in your browser using DataLab