In the second step the variables in these two working files were used to divide each individuals into pseudo-persons according to changes in category membership for the time-dependent covariates. Each record in this file had an entry date and exit date for one of the time-dependent strata and all additional information that was needed by MAS for exact computation of person-years. Table AIII shows that there were (3 x 3 x 10 = 90) possible time dependent exposure categories. Note, however, that a person can only move from lower to higher dose categories. The entry and exit dates for the cumulative external dose groups were obtained by interpolation within years after the appropriate lag had been introduced. The entry and exit dates for internal exposure groups were based on the mid-point of the year in which a category change occurred. Workers employed at K-25 or TEC before they worked at X-10 or Y-12 entered follow-up in the ``Multiple'' facility category, whereas individuals first hired at X-10 or Y-12 contributed person-years to the facility of first hire until the end of the study or until the date they became a multiple facility worker for the appropriate lag. For the situation described in Table AIII, and using a ten year lag, the 28,347 (note that 423 workers did not have at least one year of follow-up and were excluded from the analyses) workers are partitioned into 61,597 pseudo-individuals for processing by MAS. A more detailed description of the exact person-year stratification method is given in  Appendix IV) and  who also discussed other software that can be used for these computations.