Randomization Tests in Support of Some Statistical Analyses of a Weather Modification Experiment

摘要

The validities of analyses of variance and covariance applied to weather modification experimental data are examined through Monte Carlo samplings of the corresponding randomization distributions. It is found that the parametric tests approximate the randomization distributions well. Thus the use of parametric procedures in exploratory analyses of weather modification data, after appropriate transformation, is justified. The authors comment also on the issue of multiplicity of analyses in data explorations. It is pointed out that analyses that are closely related and explore inherent sources of experimental variation are unlikely to result in misleading conclusions. (Author)