Age-period-cohort analyses aim at linking the expected incidence rate Rij with the effects of the age at death Ai (age effects), the death date Pj (calendar period effects), and the birth date Ck (birth cohort effects).

Each category has the same time duration (e.g., 10 years). In this study however, an observation period was incomplete (1990-97).

These effects are considered to be multiplicative.

This is illustrated by plotting the data on a logarithmic scale (Figure 1 in Cohen CH. Am J Epidemiol 2000;152:474-9).

The age-specific incidence is considered to be stable among calendar periods of observation, and among birth cohorts.