The classification tree derived from the smoking data is illustrated in Figure 1 (left) and shows the following: From the entire sample of 200 adolescents (represented by node 1 in Figure 1 (left), where the node numbers are mere labels assigned recursively from left to right starting from the top node), a group of 92 adolescents is separated from the rest in the first split. This group (represented by node 2) is characterized by the fact that “none” of their four best friends smoke, and that within this group only few subjects intend to smoke within the next year. The remaining 108 subjects are further split into two groups (nodes 4 and 5) according to whether they drank alcohol in “one or less” or “more” occasions in the past month. These two groups again vary in the percentage of subjects who intend to smoke.
