In the third step the MAS program  was used to generate a person-years matrix for five year age and calendar year intervals for all possible combinations of factors of interest. In the example these factors are A, B, S, L, IG, F, and X (see Table AIII). Each of these matrices was then multiplied by the appropriate rate table for each cause of death category provided as input. Since we have stratified on birth cohort and age at risk the cause specific observed and expected deaths are added over the calendar year dimension (this is an option in MAS). In our example this results in a seven dimensional table that contains 4,230 cells with person years greater than zero. Each cell in the ADS contains the index values for each of the factors, the person-years, the average dose, the observed and expected deaths for each cause of death category of interest---the cause of death categories were determined by the rate table provided by . The resulting ADS (see Figure A1) for the ten year lag analyses reported in Tables VI--VIII is a SAS data set with variables named OBS j, EXP j for each value of j (i.e. cause of death category) as well as the indices corresponding to the factors A, B, S, L, IG, F, X. Each stratum in the ADS also contains the total person-years at risk (PY) and the radiation dose variable D, which is the average dose of all person-years within each stratum (calculated by a separate SAS program). The ADS used to obtain the results in Tables VI---VIII is available through CEDR .
The all cancer subset and computer programs used to obtain the results in this Appendix and Table VII can be obtained HERE