Categorical Data Anaysis.

Categorical response variables are common in many environmental observational and experimental studies; however, categorical analysis is not commonly used in many agronomic disciplines.  This presentation will briefly discuss the theory, major computational procedures, and practical suggestions for using different types of logistic regressions. Specifically, we will demonstrate examples of autologistic regressions, ordinal (proportional odds) logistic regressions, and multilevel (hierarchical) binary logistic regressions that can be used to analyze data collected in different types of on-farm evaluations of nitrogen (N) fertilizer use on corn.  These analyses were focused on several practical agronomic problems including: 1) identifying factors that can explain spatial variability in categorical economic (profitable vs not profitable) yield response to N fertilizer within spatially variable corn fields; 2) identifying management and weather factors that affect late-season corn N status; 3) classifying production fields into areas with deficient and sufficient corn N status using digital aerial imagery of the corn canopy; and 4) evaluating the calibration (N sufficiency) categories currently used with the late-season corn stalk nitrate test.