Table 5.

Efficiency^† of models for the marginal mean relative to marginalized negative binomial regression.

Parameter	n	Poisson	NB	MZIP
β 1	100	1.170	0.606	0.734
	200	1.020	0.555	0.695
	500	0.845	0.487	0.658
	1000	0.764	0.433	0.633
β 2	100	1.005	0.788	0.786
	200	0.989	0.795	0.792
	500	0.970	0.777	0.794
	1000	0.964	0.772	0.789
β 3	100	1.067	0.782	0.730
	200	0.999	0.784	0.741
	500	0.951	0.764	0.758
	1000	0.939	0.755	0.757

^†Ratio of the mean squared error (mse) of the MZINB model divided by the mse of the model in the column heading; mse is computed as the sum of the monte carlo variance of ${\widehat{\beta }}_j$ and its squared bias.