Table 3.

Simulation results for Type I error, power, and cdf estimation under 3 data-generating mechanisms and 3 models for estimation

Type I					(Y>1|X=0)	(Y>3|X=0)
True f0	Model	Error	Power	logL (SE)	Estimate (SE)	Estimate (SE)
Truncated	True	—	—	—	0.0892	0.0017
Poisson	SPGLM	0.056	0.62	– 229.3 (11.6)	0.0877 (0.017)	0.0017 (0.0007)
SPGLM	PLLM	0.055	0.61	– 231.0 (11.7)	0.0889 (0.013)	0.0018 (0.0006)
	POM	0.049	0.56	– 229.6 (11.6)	0.0907 (0.018)	0.0022 (0.0028)
0-Inflated	True	—	—	—	0.1258	0.0073
Poisson	SPGLM	0.047	0.47	– 235.3 (14.0)	0.1241 (0.021)	0.0073 (0.0024)
SPGLM	PLLM	0.091	0.58	– 245.9 (15.5)	0.0900 (0.016)	0.0018 (0.0007)
	POM	0.042	0.41	– 235.5 (14.1)	0.1281 (0.021)	0.0080 (0.0055)
0-Inflated	True	—	—	—	0.0892	0.0017
Poisson	SPGLM	0.047	0.62	– 227.7 (11.7)	0.0839 (0.017)	0.0016 (0.0007)
POM	PLLM	0.091	0.62	– 229.3 (11.7)	0.0876 (0.013)	0.0017 (0.0005)
	POM	0.042	0.66	– 227.4 (11.7)	0.0881 (0.018)	0.0016 (0.0025)

True β values are in Table 1. True f₀ values are given in Table 2. Type I error rates are equivalent under the 0-inflated Poisson and the POM DGMs because the true models are equivalent under the null hypothesis that β₁ = 0.