Marginalized mixture models for count data from multiple source populations

Journal of Statistical Distributions and Applications

Table 9 Marginalized count regression model estimates (est) and stanard errors (se) for the number of roots produced by 270 shoots of the apple cultivar Trajan

	Poisson		NB^a		MZIP		MZINB^b		MPois-Pois
Variable	est	se	est	se	est	se	est	se	est	se
Marginal mean model
Intercept	1.880	0.058	1.876	0.126	1.854	0.060	1.855	0.072	1.863	0.056
Photoperiod, 16h	-0.711	0.104	-0.706	0.188	-0.620	0.134	-0.618	0.152	-0.687	0.144
log(BAP/2.2)	0.069	0.042	0.073	0.092	0.092	0.042	0.091	0.052	0.080	0.034
Interaction	-0.176	0.077	-0.182	0.138	-0.258	0.078	-0.259	0.094	-0.218	0.075
					Zero-inflation model				Latent class mean model
Intercept					-4.262	0.732	-4.381	0.827	2.142	0.051
Photoperiod, 16h					4.159	0.753	4.264	0.846	-4.238	0.552
Mixing probability model^c
ρ									0.178	0.184
π									0.544
Model fit statistics
-2loglik	1566.4		1402.1		1250.2		1236.5		1236.4
AIC	1574.4		1412.1		1262.2		1250.5		1250.4

^aIn the NB model, \(\hat {\phi }=0.522\) (s.e. = 0.083)
^bIn the MZINB model, \(\hat {\tau }=-2.662\) (s.e. 0.351) corresponding to \(\hat {\phi }=e^{\hat {\tau }}=0.070.\)
^cIn the MPois-Pois model, π=1/(1+e ^−ρ)