植物病害量化順序尺度之優化與統計分析方法之進展

Optimizing ordinal scales for plant disease severity assessment and advances in their statistical analysis

植物病害的量化評估，對於作物病害管理、抗病育種、以及病害預警系統等資料應用與相關領域的研究至關重要。常見的病害量化評估方法包括評估病害盛行率、罹病率與罹病度。其中，過去針對罹病度的定義為病徵佔植物組織或器官的百分比，因此常評估病徵的最近似百分比（nearest percent estimate; NPE）作為罹病度的估計值。然而，並非所有評估者都能快速準確的評估NPE，因此許多研究提出可以使用量化順序尺度進行罹病度的評估。其中，Horsfall-Barratt scale即為一例，該尺度在過去已被廣泛使用在不同病害的研究上。然而，由於其非線性級距的設計，許多研究陸續提出此類尺度的設計結構容易引入誤差，進而影響後續的統計檢定力。本文回顧過去的研究，除了探討如何優化量化順序尺度的設計結構，以提升罹病度評估的準確度，同時介紹對於量化順序尺度資料統計分析方法的近期研究成果，期能使植物保護相關研究人員對此類資料的收集與分析能有進一步的認識。

Plant disease quantification is crucial for crop disease management, breeding for disease resistance, and the development of disease forecasting systems, as well as for related research and applications. Common approaches to disease quantification include evaluating disease prevalence, incidence, and severity. Traditionally, disease severity has been defined as the percentage of symptomatic area on plant tissues or organs, with the nearest percent estimate (NPE) often used as a measure. However, not all raters can quickly, precisely, and accurately assess NPE. Thus, many studies have proposed the use of quantitative ordinal scales for estimating disease severity. One example is the Horsfall-Barratt scale, which has been widely applied in studies of various plant diseases. Nonetheless, due to its nonlinear scale design, this type of scale has been criticized for introducing bias that may reduce statistical power in subsequent analyses. This article reviews past studies, discusses how to optimize the structure of quantitative ordinal scales to improve the precision and accuracy of disease severity assessments, and introduces recent advances in statistical methods for analyzing quantitative ordinal scale data, aiming to enhance researchers' understanding of the collection and analysis of such data in plant protection research.