For the internal analyses the ERR maineffects model is expressed as
where is the observed cancer death rate per
thousand person years at risk in the
cell. In this
specification all of the terms are factors (one parameter for each
level) except for age and dose, and birth cohort is the ``referent
factor''. The baseline rates are described by the first six estimates
in column 2 of Table AV , e.g. for the 1915 birth cohort the
estimated baseline rates are log (rate) = 0.728 + 5.20 * log
(age/52.5). Consequently, the estimates for the factor B in column
two (lines 2 through 6) of Table AV represent the log of the
all cancer mortality rate for each of the five birth cohorts at the
reference age 52.5 (i.e. they are the intercept parameters). The
estimates for each level of the factors S, L, IG, and F are relative
risks (in L% units), with the first level of each factor as the
referent category. The next to last estimate in column two is the
estimated dose-response coefficient for external radiation dose and
represents the change in the ERR per Sv, which is equivalent to
percent per 10 mSv (rem). Estimates of the standard errors of each of
the parameter estimates are given in column 3 of Table AV ,
and likelihood ratio based 95% CI are given for the ERR in the last
row. The 95 percent confidence intervals were obtained using the
bounds command in an AMFIT program
(AMFIT is part of epicure[11])