Skip to main content
. 2017 Oct 13;27(1):e1585. doi: 10.1002/mpr.1585

Table 6.

Ranking of the models according to the average absolute coverage error

First Simulation Study (Unadjusted Comparisons) Second Simulation Study (Adjusted Comparisons)
1. MW 0.005 1. QP 0.009
2. HC 0.008 2. NB 0.023
3. QP 0.009 3. ZINB 0.026
4. Linear 0.020 4. HC 0.092
5. ZINB 0.039 5. Linear 0.094
6. NB 0.043 6. Poisson 0.153
7. Poisson 0.193

Abbreviations: HC, heterosketasticity consistent; MW, Mann‐Whitney; NB, negative binomial; QP, quasi‐Poisson; ZINB, zero‐inflated negative binomial.

Average absolute coverage error is calculated with the absolute difference between the estimated coverage of the 95% CI (obtained from 10 000 simulations) and the target value of 0.95, averaged over the 135 settings with moderate sample sizes from the first simulation study, averaged over the 40 settings with moderate sample sizes from the second simulation study.