Modeling the distribution and incidence of bacterial wilt on tomatoesusing binary logistic models

Strategies of managing and controlling devastating bacterial wilt of tomato are investigated by fitting binary logistic models to data consisting of dummy response variable of its incidents on reported cases to the plant clinics. Binary logistic methods are used for fitting such models to data describing the distribution and incidents of bacterial wilt (BW) caused by Ralstonia solanacearum on tomato crop. The approach overcomes some of the difficulties encountered when fitting ordinary least square regression to a dummy response variable. The study used secondary data obtained from a sample of 1980 reported cases to the plant clinics. The clinic sites were purposively selected. The results of the analysis showed the BW incidents were high in Kirinyaga, Nakuru and Embu (20.6%, 20.2% and 19.1% respectively).Bacterial wilt distribution a cross counties were significant (χ2 = 202.079, p-value<0.001). Minimum relative humidity and minimum temperature were significantly influencing bacterial wilt incidents. However, agro-ecological zones and precipitation were not significant from the models. Binary logistic models estimated efficiently the incidents and distributions of BW using the backward iteration process and corresponding residual deviance and Akaike Information Criterion.