This chapter contains examples that illustrate the use of AMFIT with grouped cohort survival data. As noted earlier, AMFIT is a Poisson regression program designed specifically for use with such data. Grouping means that the data on individual members of the cohort have been cross-classified on one or more time scales and, possibly, other factors. The basic data in each cell of this cross-classification are total time at risk, which we call person-years, and the number of cases. Cells usually also contain values for additional covariates. Person-year, or event-time, tables can be made using DATAB. Making Tables with DATAB contains examples that show how to use DATAB to create event-time tables and other tables.
The first examples in this chapter use a simple artificial data set to demonstrate the basic features of AMFIT. The remaining examples illustrate the use of AMFIT with a more realistic data set constructed from the fluoroscopy data set described in Data Input and with data from a large cohort of Japanese atomic bomb survivors. The creation of the event-time table used for the fluoroscopy data examples is described in Example 8.3. The examples in this chapter also illustrate methods for hypothesis testing and the computation of confidence bounds that can be used in any of the EPICURE risk regression programs.