This chapter, a companion to Chapter 4, clarifies the logic underlying the notion of stages of drug use progression and its relation to two related models for the analysis of progressions. These log linear models were introduced in our previous research (Yamaguchi & Kandel, 1984a; Kandel, Yamaguchi, & Chen, 1992; Kandel & Yamaguchi, 1993; Yamaguchi & Kandel, 1996) and are further refined here. The models are applied to gender/ethnic differences in progression patterns among five drugs: alcohol, cigarettes, marijuana, cocaine, and heroin. The models that underlie the substantive results presented in Chapter 4 are discussed in detail in this chapter to illustrate the use of the methods.

In order for a stage of progression in drug use to be substantiated, two conditions must be met (Yamaguchi & Kandel, 1984a; 1984b). The first condition is that of a sequential order between two states. The sequential order of initiations between a lower-stage drug A and a higher-stage drug B is such that all systematic pathways of drug use progression, except random progressions, must not include the reverse sequence B-A (Yamaguchi & Kandel, 1984a). As a result, all pathways of progression other than the random type fall into one of the following three groups: (1) reaches neither stage A nor B; (2) reaches only stage A; and (3) reaches both stages A and B, stage A first. Under these conditions, the pathways of drug use progression satisfy the condition that stage A must precede stage B.