next up previous
Next: Main Effects-- Internal Up: Dose-Response Analyses (DRA) Previous: Table with Marginal

Fitting The Main Effects Model

The general Poisson regression model

was used to describe the joint effects of each of the explanatory variables of interest on cause-specific mortality. Maximum likelihood estimates of the parameters and LRT statistics were obtained using widely available software [4,11]. The baseline rate represents the age-specific death rate for individuals at the reference level of each of the explanatory variables. In previous reports on the X-10 only subcohort [15,16] a parametric model was used to describe the baseline rates, i.e.

where = age at interval mid-point. In this report we will use the external/internal model ([1], Chapter 4)

in which the baseline rates are assumed to be proportional to the known external standard rates (U. S. white male rates) was used. To contrast and further explain these two models both of them were used to summarize the data for the seven dimensional table of observed and expected deaths for all cancers for the X-10/Y-12 ever sub-cohort.