National survey of viruses in pastures of subterranean clover. II. Statistical methodology for large scale quantitative ELISA

Abstract

Statistical methodology was applied to a survey of time-course incidence of four viruses (alfalfa mosaic virus, clover yellow vein virus, subterranean clover mottle virus and subterranean clover red leaf virus) in improved pastures in southern regions of Australia, with samplings in each winter and spring over 3 years. The 100 samples per paddock collected at each time of sampling provided detection probabilities of 0.63 and 0.87 for 1% and 2% infection respectively. A microtitre plate design for ELISA was developed to include 60 field samples, 10 glasshouse-grown healthy control samples and 6 glasshouse-grown samples infected with the virus under examination. This design was used on 816 plates for each of the four viruses tested. The method used for identification of virus in sap of a plant sample from a particular paddock was that the ELISA reading was both significantly greater than healthy control readings, and an outlier in the distribution of readings of all sap samples from that paddock. It is argued that as the identification of uninfected samples as infected was highly unlikely, this double criterion method was superior to the use of each criterion separately. Use of significance above healthy control values as the sole criterion would have increased virus incidences by about 60%; use of outlier identification as the sole criterion would have increased virus incidences by about 30%. A generalized linear model with binomial errors and logit link was used for adjusting the virus incidences reported in the previous paper (Helms et al. 1993) for biases due to paddocks and/or districts not sampled on some occasions. This adjustment slightly increased overall incidences in all but one sampling and confirmed the time-course increase in incidence over the 3 years of the survey. The same model also proved to be the most appropriate for investigating the effects of year, season and district on virus incidence.