A new approach to analyse longitudinal epidemiological data with an excess of zeros

Table 2 Regression and model fit parameters for the three longitudinal models with a random intercept, evaluating the difference in hypoglycaemic events for education*

	Linear mixed model			Poisson mixed model			Two-part joint mixed model (binomial/Poisson)
	Coef.	Std. Err.	P > \|z\|	Coef.	Std. Err.	P > \|z\|	Coef.	Std. Err.	P > \|z\|
Educational level (high education)	1.14	(0.36)	0.001	0.66	(0.15)	0.000	0.62	(0.15)	0.000
BIC	11842.87			7609.073			6687.639
Mean squared residual	13.75			9.73			7.26

Abbreviation: Regression coefficients (Coef.), Standard errors (Std. Err.), P-value (P > |z|), Bayesian information criterion (BIC).
* Education is time independent, therefore random slopes could not be calculated.

ISSN: 1471-2288