Table 7.

Multivariate logistic regression models of the impact of socio-demographics and house flies knowledge on house flies preventive measures

Model	H0	−2LL*	P£
Locality + Education + Breeding site knowledge + Preventive measures knowledge + Int**		70.51	----
Locality + Education + Breeding site knowledge + Preventive measures knowledge	βInt = 0	71.78	0.866
Education + Breeding site knowledge + Preventive measures knowledge	βLocality = 0	74.74	0.261
Education + Breeding site knowledge	βPreventive measures knowledge = 0	76.76	0.043

*-2 log likelihood, ** All possible 2 way interactions, £ p values based on chi square of −2 log likelihood difference between the reduced model and initial model. The predictors: locality, education level of the farmers, farmers’ knowledge about house flies breeding sites and preventive measures, had P values < 0.25 and were the potential predictors in univariate analysis (Table 6). These four potential predictors were then entered in the multivariate model by following the methodology of Hosmer and Lemeshow [17]. In succeeding steps, the predictors with a P > 0.05 in the previous step were removed from the model until complete loss of fit (P < 0.05) of the model was achieved.