Marginalized mixture models for count data from multiple source populations

Journal of Statistical Distributions and Applications

Table 7 Marginalized count regression model estimates (est) and standard errors (SE) for the Lanarkshire caries trial

	Poisson		NB		MZIP		MZINB		MPois-Pois		MNB-Pois
Variable	est	se	est	se	est	se	est	se	est	se	est	se
Marginal mean model
Intercept	1.216	0.017	1.201	0.035	1.234	0.022	1.206	0.034	1.309	0.029	1.207	0.035
bc	0.758	0.017	0.757	0.036	0.765	0.023	0.757	0.035	0.663	0.026	0.757	0.035
calc	−0.204	0.020	−0.189	0.040	−0.211	0.025	−0.195	0.039	−0.263	0.031	−0.199	0.039
NaF	−0.072	0.018	−0.056	0.039	−0.098	0.023	−0.060	0.038	−0.144	0.028	−0.060	0.038
NaFTMP	−0.052	0.022	−0.022	0.048	−0.104	0.029	−0.034	0.047	−0.062	0.035	−0.033	0.047
					Zero-inflation model				Latent class mean model
Intercept					−1.280	0.079	−2.257	0.290	2.131	0.031	−1.425	0.513
bc					−1.309	0.113	−2.911	1.571	0.512	0.027	3.539	0.500
calc					0.129	0.095	−0.149	0.273	−0.249	0.031	0.022	0.173
NaF					0.208	0.096	0.419	0.307	−0.152	0.029	−0.013	0.188
NaFTMP					0.289	0.119	0.545	0.363	−0.081	0.037	−0.307	0.201
Mixing probability model estimates^a
ρ									−0.753	0.055	−1.818	0.170
π									0.320		0.140
Dispersion estimates^b
τ											−0.328	0.042
ϕ			0.806	0.028			0.639	0.038			0.720
Model fit statistics
−2loglik	22706		17249		20410		17189		18048		17169
AIC	22716		17261		20430		17211		18070		17193