Table 1.

Power of the inference method used by Kaul and Wolf to detect a plain packaging (PP) effect of size Δ, using pseudo data generated with normal distribution (and constant variance) and binomial distributions assuming an immediate PP effect and a gradual PP effect with the binomial. Two effects areas (see Figure 1) are considered: one defined by the “liberal” 90% confidence intervals, the other by the “more conservative” 95% confidence interval (in Kaul and Wolf’s terminology). Column 2 (with grey background) shows the values in Table 2 of Kaul and Wolf’s working paper ⁶. Power estimates were obtained with 100,000 Monte Carlo repetitions.

Δ PP Effect (%)	Power of K&W’s inference method
	Effect area based on 90% confidence intervals			Effect area based on 95% confidence intervals
	Simulation based on normal distribution, constant variance, immediate effect (K&W table 2)	Simulation based on binomial distribution		Simulation based on normal distribution, constant variance, immediate effect	Simulation based on Binomial distribution
		Immediate effect	Gradual effect		Immediate effect	Gradualeffect
(1)	(2)	(3)	(4)	(5)	(6)	(7)
0.25	0.56	0.29	0.25	0.35	0.13	0.10
0.50	0.64	0.38	0.29	0.43	0.21	0.13
0.75	0.72	0.49	0.34	0.51	0.33	0.17
1.00	0.79	0.63	0.40	0.61	0.48	0.22
1.25	0.85	0.77	0.46	0.70	0.65	0.28
1.50	0.90	0.87	0.53	0.79	0.81	0.35