Design of Experiments for Statistical Analysis of Geological Data

Abstract

Statisticians have made available a number of experimental designs based on analysis of variance techniques, some of which are particularly adaptable to geological problems in sedimentation, petrology, paleontology, and other fields. In analysis of variance, experiments are so designed, or observations are so collected, that the data permit separation of the total variability into its parts. By comparing the separate variances, it is possible to test the statistical significance of the experimental or observational data. Among the large number of analyses of variance models available, several basic forms have immediate application to a variety of geological problems. These models are described, and a geological example is shown for each. In a later section of the paper some of the assumptions underlying the method are pointed out, and a summary table lists some of the kinds of problems which may be studied with each of the models. By a slight extension of the method, analysis of variance data permit estimation of variance components, which can be used to make unbiased estimates of various geological populations. This aspect of the subject is touched upon to indicate the wide field of inquiry made available by experimental design.