Table 4.

Maximum likelihood estimates of binary logistic and probit regression models for diabetes for subjects in the Framingham offspring study in the 20-year observation period predicted by the demographic, behavioral and anthropometric variables measured at Exam 1¹

	Dependent variable: Diabetes in the 20-year period n = 3718
Independent variable	Binary logistic model				Probit model
	Coefficient	SE	95% CI for exp (β)		Coefficient	SE	Marginal effect	SE
Constant	-8.658*	0.749	-	-	-4.666*	0.373	-	-
Sex	0.289	0.172	0.952	1.870	0.153	0.085	0.013	0.007
Age	0.053*	0.008	1.038	1.072	0.026*	0.413	0.002*	0.0003
Physical activity score	0.010	0.019	0.973	1.048	0.004	0.010	0.0003	0.0008
Alcohol index	0.016	0.012	0.992	1.040	0.008	0.007	0.0007	0.0006
Cigarettes smoked	0.009*	0.004	1.000	1.018	0.005*	0.002	0.0004*	0.0002
Systolic Blood Pressure	-0.004	0.008	0.981	1.011	-0.0001	0.004	-0.0001	0.0003
Diastolic Blood Pressure	0.035*	0.012	1.012	1.060	0.016*	0.006	0.001*	0.0005
LDL	-0.001	0.002	0.995	1.003	-0.0005	0.001	0.00004	0.0009
HDL	-0.038*	0.006	0.951	0.975	-0.018*	0.003	-0.002*	0.0003
BMI	0.112*	0.016	1.085	1.154	0.059*	0.008	0.005*	0.007
Chi-square statistic2, df = 1	0.005	-			1.29	-

¹ Values are slope coefficients and standard errors; see equation (4) in the text. ² Chi-square test for combining height and weight as the BMI based on a logarithmic specification of the model. * P < 0.05.