This is an artificial data set similar (but not exactly equal! [*]) to the one described in Breiman et al. (1984, p.238). The cases are generated using the following method:
Generate the values of the 10 attributes independently using the following probabilities:

Obtain the value of the target variable Y using the rule:

[*] - Thanks to Nitin Indurkhya to pointing out that the data set I'm using it is not exactly equal to the one on the CART book. The later uses a gaussian error component with variance 2, while I'm using a variance of 1.