Montgomery and Peck (1982) illustrated variable selection techniques on the Hald cement data and gave several references to other analysis. The response variable y is the heat evolved in a cement mix. The four explanatory variables are ingredients of the mix, i.e., x1: tricalcium aluminate, x2: tricalcium silicate, x3: tetracalcium alumino ferrite, x4: dicalcium silicate. An important feature of these data is that the variables x1 and x3 are highly correlated (corr(x1,x3)=-0.824), as well as the variables x2 and x4 (with corr(x2,x4)=-0.975). Thus we should expect any subset of (x1,x2,x3,x4) that includes one variable from highly correlated pair to do as any subset that also includes the other member.
cavendish
(Package: wle) :
Cavendish's determinations of the mean density of the earth Data
The Cavendish's determinations of the mean density of the earth data (relative to that of water) are 29 measures performed in the 1798 using a torsion balance devised earlier by Michell. After the sixth of these determinations, Cavendish changed his experimental apparatus by replacing a suspension wire by one that was stiffer. To further complicate matters, Cavendish erred in taking the mean of al 29 determinations by treating the value 4.88 as if it were in fact 5.88.
This data set consists of 60 observations for two variables (ydata and xdata). The appropriate regression model for the first fiftyfourth obsertvations should be like y = x^2 + e the last sixth comes from a different model.
artificial
(Package: wle) :
Hawkins, Bradu, Kass's Artificial Data
This data set was generated by Hawkins, Bradu and Kass in the 1984 for illustrating some of the merits of a robust technique. The data set consists of 75 observations in four dimensions (one response and three explanatory variables). The first 10 observations are bad leverage points, and the next four points are good leverage points (i.e., their x are outlying, but the corresponding y fit the model quite well).
providing 
.