INFEST: an R web application to perform statistical analysis of insect feeding behavior data from EPG systems

Eletropenetrography (EPG) can help to unveil the feeding behavior of sucking insects feeding from a host of interest, including vegetal and animal tissues. It can be used to evaluate the effect of treatments such as insecticides. Based on a user's interpretation of electrical waveforms, a single EPG experiment can generate dozens of response variables associated with specific feeding activities. The variables consist, mostly, of duration and count of events. Some variables often present problematic data, such as an excess of zeros and overdispersion, which causes the classical statistical methods to be ineffective in discriminating treatments. In this article, we present INFEST-Insect Feeding Behavior Statistics, a user-friendly web application to process raw data files and to fit sophisticated statistical models to perform statistical analysis of EPG data. The program can import files from DC (direct current) and AC-DC (alternating current-direct current) EPG monitors' software and automatically calculates duration, count, and sequential variables. The statistical approach to compare experimental groups is based on Generalized Additive Models for Location, Shape, and Scale-GAMLSS, a very flexible class of parametric models that incorporates several probability distributions, including normal.