Table 3. Parameter estimates, standard errors (SE) and p-values for the zero-inflated Negative Binomial model under the latent formulation (with derived overall effects) and the marginal log-logit model with direct overall effects.

	Latent log-logit model with derived overall effects on marginal∓			Marginal log-logit model (direct overall effects)
Effects	Estimate	SE	p-value	Estimate	SE	p-value
Mean
Intercept	1.915 (1.084)	0.098 (0.121)	< 10−4 (< 10−4)	1.278	0.104	< 10−4
Unemployed	0.175 (0.397)	0.105 (0.003)	0.096 (0.295)	0.295	0.110	0.007
Age	0.268 (0.974)	0.071 (0.093)	< 10−3 (< 10−4)	0.700	0.077	< 10−4
SI	0.056 (0.265)	0.074 (0.085)	0.450 (0.002)	0.266	0.081	0.001
PHP	0.136 (0.109)	0.049 (0.050)	0.006 (0.029)	0.128	0.049	0.009
Age*SI	-0.057 (-0.347)	0.077 (0.091)	0.460 (< 10−3)	-0.258	0.081	0.001
Susceptibility probability
Intercept	0.331	0.213	0.120	0.228	0.193	0.250
Unemployed	0.497	0.232	0.032	0.572	0.210	0.006
Age	1.817	0.254	< 10−4	1.635	0.208	< 10−4
SI	0.200	0.160	0.211	0.241	0.142	0.090
Age*SI	-0.254	0.211	0.229	-0.238	0.182	0.191
Dispersion log(κ)
Intercept	-0.005	0.115	0.965	-0.015	0.114	0.895
Summary statistics
Max logL		-1872.4			-1877.5
AIC		3768.8			3779.0

^∓Parameter estimates and inferences for the derived overall effects on the marginal mean under the latent log-logit model are in parentheses.