Link to home

Information Graphs for Binary Predictors

January 2015 , Volume 105 , Number  1
Pages  9 - 17

G. Hughes, N. McRoberts, and F. J. Burnett

First and third authors: Crop and Soil Systems Research Group, SRUC, The King's Buildings, West Mains Road, Edinburgh EH9 3JG, UK; and second author: Plant Pathology Department, University of California, Davis, CA 95616-8751.

Go to article:
Accepted for publication 24 June 2014.

Binary predictors are used in a wide range of crop protection decision-making applications. Such predictors provide a simple analytical apparatus for the formulation of evidence related to risk factors, for use in the process of Bayesian updating of probabilities of crop disease. For diagrammatic interpretation of diagnostic probabilities, the receiver operating characteristic is available. Here, we view binary predictors from the perspective of diagnostic information. After a brief introduction to the basic information theoretic concepts of entropy and expected mutual information, we use an example data set to provide diagrammatic interpretations of expected mutual information, relative entropy, information inaccuracy, information updating, and specific information. Our information graphs also illustrate correspondences between diagnostic information and diagnostic probabilities.

Additional keywords: diagnosis, disease management, entropy, information theory.

© 2015 The American Phytopathological Society