Assessing disease severity: accuracy and reliability of rater estimates in relation to number of diagrams in a standard area diagram set

Abstract

Error in estimates of plant disease severity occur and standard area diagrams (SADs) help improve accuracy and reliability. The effects of diagram number in SADs are unknown. The objective of this study was to compare estimates of pecan scab severity made without SADs, and using three‐, five‐, seven‐ or 10‐diagram SADs. Disease severity was estimated to the nearest percent (NPE), or classified to the closest reference diagram value using a scale. Twelve raters assessed 20 images of scabbed pecan valves with and without the SADs using NPEs and the scale method (values were converted to midpoints prior to analysis). Increases in diagram number using NPEs did not necessarily result in more accurate or reliable estimates. Inter‐rater reliability was positively correlated with number of diagrams using NPEs (r = 0·3288 (P < 0·0001)) or a scale (r = 0·2803 (P < 0·0001)). The least accurate estimates improved the most using SADs with NPEs, but the gain did not relate to number of diagrams; as few as three diagrams reduced error of inaccurate estimates as much as five, seven or 10 diagrams. When used as a scale, only estimates made with 10‐diagram SADs had similar accuracy and inter‐rater reliability to estimates made using SADs with NPEs. Maximum disease severity and the disease severity range for the pathosystem and the SADs are important factors and will probably influence resulting accuracy and reliability. The ramifications of diagram numbers in SADs and assessment methods on accuracy and reliability of disease estimates are discussed.