Table 2
Estimates of bias in cumulative exposure-disease trends derived via Poisson regression under four rules for exposure categorisation and three methods for exposure score assignment; simulation analyses under the assumption that exposure intensity conforms to a uniform distribution
| Exposure categorisation | Score assignment | ||
|---|---|---|---|
| Category midpoint | Category mean | Cell specific mean | |
| Bias, λ | Bias, λ | Bias, λ | |
Coefficient of bias, . Estimates were derived after 100 repetitions of each simulation. | |||
| 5 categories | 1.01 | 1.01 | 1.00 |
| Based on quintiles of the cumulative exposure distribution | |||
| 10 categories | 0.99 | 1.00 | 0.99 |
| Based on deciles of the cumulative exposure distribution | |||
| 5 categories | 0.97 | 1.00 | 0.99 |
| Based on quintiles of the cumulative exposure distribution among cases | |||
| 10 categories | 0.98 | 1.00 | 0.99 |
| Based on deciles of the cumulative exposure distribution among cases | |||