Revisiting the use of disease index and of disease scores in plant pathology

Abstract

Plant disease assessment often involves the use of rating scores e.g., “0, 1, 2, … 6”, to rapidly record disease severity. Scores are ordinal, but qualitative, variables. The analysis of scores with parametric statistical methods will unavoidably lead to confusion and error. By the same token, using disease scores to calculate sum, mean, or derived indices such as a disease index, is incorrect. The proper analysis of disease data recorded as scores depend on the disease assessment key used to define the disease scores. If disease scores are categorical variables describing symptom attributes, they should be analysed using non-parametric methods for qualitative variables. If disease scores represent disease severity as the fraction of the diseased area, there are two ways for analysis. First, if disease scores correspond to a pre-set number of disease severity values, scores should be back-transformed to these severity values. But if disease scores correspond to successive severity intervals, scores should be converted into severities at the mid-range of these intervals. In both latter cases, the new variables derived from these transformations are continuous, quantitative variables. They may therefore be analysed with parametric methods. Numerical examples are provided to illustrate the outcomes of different ways to process disease rating score data.