The author disentangles the age sex, spatial, and temporal structure of mortality in the Czech Republic during the period 1987-97 for a selection of cause-of-death categories and investigates possible socioeconomic and other causes for the mortality differences uncovered. The clarification of the major effects, as well as possible interactions between them, verifies whether causes of death should be modelled separately, and which interactions should be tested in the explanatory analysis. For instance, it was considered inappropriate to incorporate district-level time-series data in the analysis when spatial mortality differences did not show significant variation between 1987 and 1997-as occurred with cancer and digestive system diseases. In this case, the exogenous variables would take on the average for the study period. © 2008 Pion Ltd and its Licensors.