Relevance of Epidemiology to Identifying Huanglongbing Resistance: The Power of Experimental Design

Abstract

Biological phenomena are influenced by numerous factors and interactions. As such, their observation as affected by different treatments often takes on a distribution of responses, the perceived form of which depends on aspects of experimental design. If sampling sizes or replicates are too few, then misleading conclusions may ensue, since relatively limited data present only a slice of the full range of responses that, under varying conditions, an individual treatment might contribute toward. An analysis involving simulated subsampling of actual huanglongbing data was conducted to illustrate the effect of varying sample sizes and replicates on results. At one end of the spectrum, increased sample sizes (while maintaining only one replicate) increased the rate of significantly different (α = 0.05) estimates of disease incidence under one treatment as compared to the control (complete sampling: 150 trees per treatment, three replicates) to ~33%. Conversely, with a fixed per treatment sample size of 10 trees, estimates of disease incidence were respectively up to 75%, 40%, or 25% different from complete sampling estimates when one, two, or three replicates were utilized. Thus, too few replicates or too few samples per replicate can lead an investigator to infer apparent differences among treatments when larger sample sizes and/or more replicates would demonstrate a lack of statistical difference. Though the analyzed data were based on the effects of various control strategies on development of huanglongbing disease incidence, results are analogously applicable toward alternative investigations, such as evaluation of resistant lines.