Effects of Rater Bias and Assessment Method on Disease Severity Estimation for Comparing Treatments by Using Meta-Analysis Approach

• Main Authors: I-Hsuan Lee, 李易軒
• Other Authors: Kuo-Szu Chiang
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/60678723586039873476

碩士 === 國立中興大學 === 農藝學系所 === 104 === Abstract The data of plant disease severity is widely used in many agricultural studies, such as predicting yield loss, comparing treatments, assessing crop germplasm for disease resistance, and monitoring and forecasting of disease outbreaks. Visual estimation of disease severity as a proportion of area often used in studies of plant pathogy and plant breeding; however, it is error prone. The purpose of this study was to investigate the effects of rating bias and assessment scale method on comparing treatments. In order to estimate the effect of overestimation on the real situation, Meta-Analysis method was employed to combine the performance of different raters. The data used in the study was investigated on citrus canker (Xanthomonas citri subsp. citri) from grapefruit trees in south Florida in the USA. Twenty-eight raters joined the study and all 200 leaves assessed for each rater. From the results of the study, the power of the hypothesis tests (comparing treatments) using unbiased estimates was most great and the next better is the power of absolutely overestimated raters. The power of partially overestimated raters are the worst. Moreover, there are little difference among different assessment methods. Only in severity of 20% was the power of the hypothesis using H-B inferior to the other methods. Also, there was lower power for low actual severities in linear 10% scale. Finally, the results of this study could be helpful in severity assessment of plant diseases.