Figure 1

Variance vs mean of Q. This scatter plot of Var[Q]/(K−1) vs. E[Q]/(K−1) contains the results of simulations of the moments of Q _LOR for the 144 configurations of parameters: K= 5, 10, 20, 40; N= 90, 150, divided equally into the two arms; log odds ratios: 0, 0.5, 1, 1.5, 2, 3; and control probabilities: 0.1, 0.2, 0.4. The studies in each simulation all have the same parameters. The simulations for each configuration were replicated 10,000 times. The grey reference line (Var[Q]=2E[Q]) indicates the relation that would be expected if Q followed a chi-square distribution. (a): N=90 black and N=150 red. (b): K=5 (black), K=10 (red), K=20 (blue) and K=40 (green). The black curve corresponds to the fitted quadratic equation Var[Q _LOR ]/(K−1)=4.74−12.17E[Q _LOR ]/(K−1)+9.42[E[Q _LOR ]/(K−1)]².