Next: Dose-Response Analyses. Up: Statistical Methods Previous: Grouped Data Poisson

### Facility Comparison Analyses.

Facility is the risk factor of primary interest. Mortality differences between work-forces at these facilities reflect a combination of occupational and other factors. A detailed discussion of the potential for exposure to internal and external radiation and other related hazards at each facility is given by Watkins et al [2,3]. In these analyses an evaluation of differences of cause specific mortality among the Oak Ridge plants is emphasized.

A preliminary analysis of death rates due to all causes with three explanatory variables (birth cohort, age at risk, and facility) is presented to demonstrate the relationship between the two approaches that are used to "adjust" for age at risk.

The rest of the results are based on Eq. 1 with a multiplicative main effects model using external age-cause specific death rates (from U.S. vital statistics) to compute expected deaths. The resulting ADS contains the observed deaths, the ''expected" deaths, and the level of each of the factors (which is equivalent to a covariate vector) SES, calendar period, length of employment, and facility. The multiplicative main effects model is expressed on a logarithmic scale as

For convenience in describing results the convention---[13] Chapter 22---of dropping Greek letters (that represent the unknown parameters) and listing the explanatory variables that define the relative risk function is used.

This corresponds to standard GLIM notation [14] for a log-linear model in which F is the facility factor (5 levels), S is the socioeconomic status proxy paycode(2 levels), L is the length of employment factor (2 levels), and t = (calendar year - 1965)/100, which represents a calendar period trend in the log of the SMR. The indicator variables for the factors F, S, and L are coded so that the coefficients for the facilities represent the log of the SMR for each facility at the reference level of S and L when t = 0 (i.e. in 1965 which is the midpoint of follow-up). The coefficient for t describes the change in the log SMR over follow- up. The results of fitting the main effects model for selected causes of death by race gender groups are presented in tabular form.

Next: Dose-Response Analyses. Up: Statistical Methods Previous: Grouped Data Poisson