Log-linear based estimates of prevalent count of subjects with alcohol related problems. Models include interaction terms involved to the adjustment of estimates for dependence among sources. Summary of forward selection strategy of the best model (Voghera, northern Italy, 1997)
| Model | df3-151 | D3-151 | Number (95% CI)3-152 |
| (1) Main effects only (u1+u2+u3+u4) | 103-151 | 31.1333-150 | 1492 (1121, 2082) |
| (2) Main effects + 1 two factor interaction (u1+u2+u3+u4+u23) | 9 | 16.7183-150 | 1818 (1306, 2703) |
| (3) Main effects + 2 two factor interactions (u1+u2+u3+u4+u23+u34) | 8 | 10.3433-150 | 2297 (1524, 3794) |
| (4) Main effects + 3 two factor interactions (u1+u2+u3+u4+u23+u34+u14) | 7 | 8.8303-150 | 1845 (1164, 3367) |
| (5) Main effects + 2 two factor interactions + 1 three-factor interaction (u1+u2+u3+u4+u23+u34+u234) | 7 | 8.9763-150 | 2813 (1487, 4564) |
↵3-151 Degrees of freedom of the model and corresponding residual deviance.
↵3-152 Number of subjects with alcohol related problems estimated according to the parameters of the log-linear model and corresponding 95% confidence intervals.
↵3-150 p<0.05; the tested null hypothesis is that the model fits the data; alternative hypothesis is a significant departure of the model from the data. ui: parameter of the main effect of the i-th flagging source (i = 1,..,4); uij: parameter of the two factor interaction between the i-th and the j-th flagging sources (i = 1,..,4; j = 1,..,4); uijk: parameter of the three factor interaction between the i-th, the j-th and the k-th flagging sources (i = 1,..,4; j = 1,..,4; k = 1,..,4); Flagging sources: F1: self help volunteering groups (anonymous alcoholics and on treatment alcoholic groups); F2: Psychiatric Ambulatory; F3: Public Alcohology Service; F4: hospital discharges.