Chapter 10 Robust Regression 487 computes the TS regression estimator just described. When p > 1, the default maximum number of iterations when using the Gauss­Seidel method, indicated by the argument iter, is 10. The function correg(x,y,corfun=tau) computes the Theil­Sen estimate by determining b 1 , . . . , b p so that | j | is approximately ^ equal to zero, where j is Kendall's tau between the jth predictor, x j , and ^ y - b 1 x 1 - · · · - b p x p . It generalizes the function tsreg by allowing Kendall's tau to be replaced by some other correlation. For example, correg(x,y,corfun=pbcor) would use the percentage bend correlation coefficient. When dealing with a single predictor, only two R commands are needed to create a scatterplot that includes a regression line. (The two R functions are plot and abline.) To make this task even easier, the R function regplot(x,y,regfun=tsreg,xlab="X",ylab="Y") is supplied, which by default plots the Theil­Sen regression line. Other regression lines can be