Sample size (N) needed to detect a significant association between a risk factor and injury for some chosen values of log odds ratio (β) and injury frequency (p). Assumption: Univariate Cox regression model with censorship probability 1-q=0.10. This means that on the average the non-injured subjects in the study will be absent for other reasons than injury for 10% of the observation period
| Frequency (p) | LogRR (β) | RR | Sample size (n) | Projected number of injured subjects (n • p) |
|---|---|---|---|---|
| 0.05 | 0.25 | 1.28 | 3929 | 196 |
| 0.50 | 1.65 | 982 | 49 | |
| 0.75 | 2.12 | 437 | 22 | |
| 0.10 | 0.25 | 1.28 | 2074 | 207 |
| 0.50 | 1.65 | 518 | 52 | |
| 0.75 | 2.12 | 230 | 23 | |
| 0.15 | 0.25 | 1.28 | 1464 | 220 |
| 0.50 | 1.65 | 366 | 55 | |
| 0.75 | 2.12 | 163 | 25 | |
| 0.20 | 0.25 | 1.28 | 1166 | 233 |
| 0.50 | 1.65 | 292 | 58 | |
| 0.75 | 2.12 | 130 | 26 |