The Influence of Variation among Replicates on Repeatability
This chapter first illustrates how a large number of observations can inform the design of a particular experiment intended to test rigorously the causes of the observed associations. It does so through the discussion and analysis of two case studies drawn from experimental immunology: (i) the efficacy of vaccines for Newcastle disease virus in poultry, and (ii) cancer immunotherapy. It then shows that it is important when repeating experiments to reproduce not only the expected result (i.e., the estimated mean value) but also to reduce the variability in estimates of the mean. Reduction in overall variance can be accomplished both with more precise measurements and by accounting for additional sources of error (covariates or “nuisance” variables). When potentially important covariates are recorded during an experiment, they can be included in the analysis of the data and help to isolate the true “signal” of the experimental treatment from the “noise” of the experimental environment.