Jeffrey L. Willers, USDA-ARS, Genetics and Precision Agriculture Research Unit, Mississippi State, MS 39762, Yongfeng Zhao, Department of Statistics, Drawer M A, Mississippi State, MS 39762, Jixiang Wu, Mississippi State University, USDA-ARS P.O.Box 5367, Mississippi State, MS 39762, Kenneth Hood, Hood Farms & Gin, 154 Perthshire Rd, Gunnison, MS 38746-9516, and John R. Bassie, Bassie Ag Service, 110 Thomas St., Cleveland, MS 38732.
Geo-referenced field samples for the Tarnished Plant Bug (Lygus lineolaris Miridae:Heteroptera) from the third week of July during the 2004 cotton production season are analyzed by several statistical methods. Geo-referenced imagery of a complex of adjoining cotton fields was processed into a map and used by two different samplers to select sample locations. An inappropriate analysis of these data describing the correlation between sample counts and imagery classes is first presented. This is followed by an iterative sequence of two appropriate analysis methods based upon generalized linear models and resampling concepts. With these methods, the sample counts are modeled as functions of categorical explanatory variables. It is argued that the generalized linear model approach based upon Poisson regression is a valuable tool for the research or industrial investigator while the resampling method is preferred for farm applications.
Audio File