VIEW ARTICLE

Disease Detection and Losses

Accuracy, Precision, and Correlation to Yield Loss of Disease Severity Scales for Corky Root of Lettuce. R. Douglas O’Brien, Department of Plant Pathology, University of California, Davis 95616; Ariena H. C. van Bruggen, Department of Plant Pathology, University of California, Davis 95616. Phytopathology 82:91-96. Accepted for publication 19 August 1991. Copyright 1992 The American Phytopathological Society. DOI: 10.1094/Phyto-82-91.

A method for comparing disease severity scales using accuracy, precision, and correlation to yield loss was tested using disease assessment scales developed for corky root of lettuce. Two qualitative interval scales, a seven-level scale developed for assessing corky root severity of mature lettuce plants and a ten-level scale developed for screening seedlings for resistance 1 mo after inoculation, were compared with a 12-level Horsfall-Barratt (H-B) scale based on the percentage of the taproot area with corky root symptoms. To estimate accuracy and precision, six taproots in each of three severity classes (0–20, 20–80, and 80–100% of taproot showing corkiness) were rated by four plant pathologists who had experience assessing corky root disease, three general plant pathologists, and three novices using each of the scales. Estimates of accuracy and precision associated with each scale were made by calculating the bias and standard deviation, respectively, divided by a correction factor reflecting the number of intervals in each scale. Correlations between yield loss and severity score for each scale were compared using severity measurements from four corky root epidemics. No single scale was identified as the best for all situations. The qualitative scale for mature plants was generally the most accurate and most precise, but the H-B scale was most accurate for roots in the 20–80% severity class. Severity scores using the H-B scale correlated best with yield loss. The qualitative scale for seedlings was inferior to the other scales except for early season yield loss prediction. Novices assessed disease severity with equally high bias regardless of the scale used, whereas plant pathologists (with or without experience with corky root) were less biased when using two of the three scales. Correlations between disease severity and yield loss varied with lettuce phenological stage and severity scale used.