Table 1.
Simulation results comparing the proposed cSIR method with the standard cancer cluster SIR estimation method (CDC) and a similar Poisson regression approach (PR). Shown are the bias in the point estimate, the coverage rate of the true SIR for 95% confidence/credible intervals (Coverage), and the width of the 95% confidence/credible intervals (CI Width).
| True SIR | Method | Bias | Coverage | CI Width | |
|---|---|---|---|---|---|
| Simulation 1 | 1 | CDC | −0.00 | 0.96 | 0.70 |
| PR | 0.20 | 0.99 | 4.81 | ||
| cSIR | −0.02 | 0.94 | 0.66 | ||
| Simulation 2 | 1 | CDC | 0.55 | 0.13 | 0.79 |
| PR | −0.27 | 0.99 | 5.86 | ||
| cSIR | −0.00 | 0.95 | 0.50 | ||
| Simulation 3 | 1.5 | CDC | −0.54 | 0.09 | 0.58 |
| PR | 1.23 | 1.00 | 49.08 | ||
| cSIR | −0.05 | 0.94 | 0.87 | ||
| Simulation 4 | 1.5 | CDC | 0.51 | 0.26 | 0.83 |
| PR | 3.76 | 0.99 | 112.96 | ||
| CSIR | −0.01 | 0.94 | 0.62 | ||