Using survival analysis for quantitative ordinal scale plant disease severity data

Authors: CHIANG, KUO-SZU - National Chung-Hsing University; LUI, HUNG - National Chung-Hsing University; Bock, Clive

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 5/13/2020
Publication Date: N/A
Citation: N/A

Interpretive Summary: Abstract Only.

Technical Abstract: In plant pathology, disease severity is typically defined as the proportion of a plant or plant organ showing symptoms of the disease. Generally, a numeric scale is used to quantify the severity, providing a standardized, convenient and rapid method for rating. ‘Quantitative ordinal scales’ (QOS) are scales which divide the percentage scale into a set number of intervals. There are different ways to process and analyze these ordinal-based interval data. Traditionally, the interval is replaced by mid-point imputation or by the class number of the interval. However, such interval based data may not be precise enough. In this situation, where we know only that a value is within a particular range, the data may be interval-censored. The body of techniques designed to deal with censoring is known as “survival analysis” (SA). SA uses all available information within a parametric modeling framework, and takes into account the uncertainty due to censoring. In order to explore the advantages and disadvantages of SA using a QOS, we implemented a simulation method. Based on the power of the hypothesis test, the results indicate that, when comparing treatments, SA had superior performance under most conditions as compared with mid-point imputation or class number. We conclude that SA is a useful option to reduce the risk of type II error when analyzing QOS severity data.