Binomial regression models for spring and summer infestations of the Russian wheat aphid (Homoptera: Aphididae) in the Southern and Western Plains States and Rocky Mountain region of the United States

Counts of Russian wheat aphid, Diuraphis noxia (Mordvilko), on winter and spring wheat, Triticum vulgare Vill. aestivum L., and spring barley, Hordeum vulgare L, from six states were described by the negative binomial distribution. These data were fitted to develop three negative binomial-based binomial regression models. Results indicated that the negative binomial model that relied upon a kmle − X¯ (average Russian wheat aphids per tiller) relationship (maximum likelihood model) minimized prediction error for P 1 the proportion of infested tillers, and X¯ in two of three crops whereas models that relied upon Taylor’s power law or the k2 − X¯ relation minimized prediction errors in just one crop. When prediction errors from the Nachman and logit models were compared with the best NBD-based models, the logit model consistently minimized prediction error for PI whereas the Nachman model minimized prediction error for X on winter wheat. The maximum likelihood model minimized prediction error for X¯ in spring wheat and spring barley. These models are suggested for use when predicting P 1 from X¯ or vice versa. An iterative algorithm is provided for predicting X with the maximum likelihood model.