Table 2:

Percent relative bias for estimation of p_z, the zero-count experiment rate as a percentage of observed experiments, for Negative Binomial and Delaporte models as obtained from 1,000 simulated datasets. Parameter μ is the expected number of foci per experiment, ϕ, σ and ν are additional scale and shape parameters. Negative Binomial performs well and, while Delaporte often underestimated p_z, with at least 1,000 contrasts it always has bias less than 10% (positive bias over-estimates the file drawer problem).

Negative Binomial
	Parameter values			% relative bias of ${\widehat{\mathbf{p}}}_{\mathrm{z}}$
	μ	ϕ	pz	$\mathbb{E}[\mathrm{k}]=200$	500	1000	2000
	4	0.4	62.1	8.76	2.85	1.40	0.80
	4	0.8	31.3	2.72	1.97	−0.78	0.32
	4	1.0	25.0	1.70	1.21	0.72	0.16
	4	1.5	16.6	2.85	1.66	0.71	0.13
	8	0.4	42.0	7.07	3.17	0.70	−0.22
	8	0.8	17.2	4.35	1.57	0.49	0.21
	8	1.0	12.5	3.32	0.84	0.63	0.04
	8	1.5	6.7	2.53	0.90	0.69	0.21
Delaporte
Parameter values				% relative bias of ${\widehat{\mathbf{p}}}_{\mathrm{z}}$
μ	σ	ν	pz	$\mathbb{E}[\mathrm{k}]=200$	500	1000	2000
4	0.5	0.02	11.8	−12.65	−10.66	−9.69	−8.02
4	0.9	0.02	20.8	−17.47	−12.35	−10.28	−10.10
4	1.2	0.02	27.6	−20.46	−18.59	−16.64	−13.83
4	0.5	0.06	10.5	−6.13	−5.58	−4.77	−3.73
4	0.9	0.06	18.0	−4.08	−4.18	−1.32	0.15
4	1.2	0.06	23.4	−5.07	−3.27	−1.09	0.01
4	0.5	0.10	9.3	−4.53	−3.32	−2.65	−1.90
4	0.9	0.10	15.6	1.07	3.86	1.91	1.80
4	1.2	0.10	20.0	4.46	3.32	3.89	4.07
8	0.5	0.02	3.6	−13.77	−10.02	−7.75	−5.91
8	0.9	0.02	9.2	−11.99	−9.00	−7.46	−5.04
8	1.2	0.02	13.8	−14.07	−11.22	−9.48	−8.12
8	0.5	0.06	2.8	−3.04	−3.18	−2.12	−1.89
8	0.9	0.06	6.8	1.41	4.13	2.74	2.47
8	1.2	0.06	10.0	10.49	7.52	7.91	5.19
8	0.5	0.10	2.2	0.93	1.36	0.82	0.74
8	0.9	0.10	5.0	8.09	5.88	5.78	4.94
8	1.2	0.10	7.3	17.91	10.68	7.77	4.76