Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: May 24, 2009
Publication Date: August 16, 2009
Citation: Wraight, S.P., Jaronski, S., Ramos, M., Griggs, M., Vandenberg, J.D. 2009. Statistical considerations in the analysis of data from replicated bioassays. Meeting Abstract. 42:82. Technical Abstract: Multiple-dose bioassay is generally the preferred method for characterizing virulence of insect pathogens. Linear regression of probit mortality on log dose enables estimation of LD50/LC50 and slope, the latter having substantial effect on LD90/95s (doses of considerable interest in pest management). Susceptibility of arthropods to pathogens varies markedly with time, resulting in high between-assay variability. This can become problematic, as assays are more time and resource intensive than single-dose tests, and experiments involving more than a few treatments may necessitate testing over time. In addition, because of the inherently high variability of assays, demonstration that results are repeatable is desirable and often demanded by journal editors and reviewers. Our findings are from replicated bioassays of the entomopathogenic fungus Beauveria bassiana strain GHA against larval hemipteran, lepidopteran, coleopteran, and orthopteran hosts and include observations on the distributions and variability of commonly used bioassay statistics. Results indicate that replicated estimates of lethal doses, slopes, and lethal dose ratios from probit/logit regression analyses of repeated bioassays are amenable to traditional statistical analyses, including parametric ANOVA, an approach rarely discussed in general treatises on analysis of bioassay data, which tend to focus almost exclusively on within-assay variability (assay precision). Application of conventional ANOVA in analyzing statistics from probit/logit analyses provides unbiased estimates of mean lethal doses and slopes with realistic standard errors reflecting the full variability of host-pathogen interactions and offers many options for rigorous multiple comparisons.