DOMESTIC, EXOTIC, AND EMERGING DISEASES OF CITRUS, VEGETABLES, AND ORNAMENTALS (DEED)
Location: Subtropical Plant Pathology Research
Title: Disadvantages of the Horsfall-Barratt Scale for estimating severity of citrus canker
| Bock, C. H. - |
| Parker, P. E. - |
| Ferrandino, F. - |
| Welham, S. - |
| Van Den Bosch, F. - |
| Parnell, S. - |
Submitted to: American Phytopathological Society Annual Meeting
Publication Type: Proceedings
Publication Acceptance Date: April 25, 2009
Publication Date: June 1, 2009
Citation: Bock, C., Gottwald, T.R., Parker, P., Ferrandino, F., Welham, S., Van Den Bosch, F., Parnell, S. 2009. Disadvantages of the Horsfall-Barratt Scale for estimating severity of citrus canker. American Phytopathological Society Annual Meeting. 98:S23.
Direct visual estimation of disease severity to the nearest percent was compared to using the Horsfall-Barratt (H-B) scale. Data from a simulation model designed to sample two diseased populations were used to investigate the probability of the two methods to reject a null hypothesis (H0) using a t-test. The relative closeness of the means, their standard deviation and rater ability all influenced the probability of Type II error (failure to reject H0). If differences were gross between means or if raters were very imprecise then both methods had similar probability to reject H0. If populations had the same mean, there was no difference in ability of the two methods to accept H0. However, if these criteria were not met, the H-B scale had higher probability of Type II error, particularly at severities 20-50%. Larger sample size reduced the probability of Type II error by both methods, but invariably the sample size required to reject H0 at P=1.0 was smaller for direct estimation. Approx. 50% more samples were required for the H-B scaled data to reject H0 with the same probability as directly estimated data. The structure of the H-B scale predisposes hypothesis testing to greater risk of Type II error compared to direct estimation. The H-B scale, and other scales, can save time, and are easy to learn and apply, but if used, sufficient sample size is needed to minimise risk of Type II errors. Direct estimation using the 1-100% ratio scale is preferable when assessing plant disease.