Learning from Yield Monitors: A Bayesian Approach
J. W. Hopkins, Gary Schnitkey, Mario J. Miranda, Luther Gilbert Tweeten · ASSA, CSSA and SSSA · 1999
A model of optimal nitrogen applications in the face of uncertain Yield response is presented. Candidate yield responses are taken from university recommendations for corn in Michigan, Ohio, and Indiana. Fertilization rates are determined by producer beliefs about whether the plot is low or high yielding. Data from a yield monitor is combined with producer beliefs using Bayes' Rule, resulting in updated beliefs. Learning is the process of updating beliefs and reducing uncertainty about yield response. Two types of learning strategies are presented-passive and active learning. Active learning allows for a higher probability of discovering the true yield response function as well as faster learning in general.