Balancing precision and power in determining the number of raters needed for standard area diagram validation studies.

Résumé fourni par la source

Abstract Visual estimation of plant disease severity is widely used but inherently subjective, which results in variability among raters. Standard area diagrams (SADs) are commonly employed to improve accuracy and precision; however, the number of raters required to validate SAD performance is often arbitrarily defined. This study aimed to establish an empirical framework to determine the minimum number of raters needed to achieve both reliable agreement estimates and sufficient statistical power to detect improvements due to SAD use. Data from nine SAD validation studies across diverse plant disease pathosystems were analyzed. Agreement was quantified using Lin's concordance correlation coefficient (LCCC). Two complementary criteria were applied: a precision-based approach using the coefficient of variation (CV) of rater agreement, and a power-based approach to detect a predefined improvement (ΔLCCC = 0.10) between unaided and aided assessments. Results showed that the precision criterion required relatively few raters (pooled minimum = 7), whereas the power-based criterion was more demanding, with final recommendations ranging from 4 to 20 raters depending on disease complexity. Pathosystems with visually complex symptoms required larger sample sizes due to greater inter-rater variability. This framework provides evidence-based guidance to optimize experimental design, improving efficiency and reproducibility in SAD validation studies. An interactive web application was developed to support adaptive estimation of rater numbers.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.