On the presentation of statistical analysis: reason or ritual

Abstract

This paper addresses the shortcomings in the presentation of statistical analyses appearing in this and similar journals. The role of statistical analysis is, in general terms, to provide criteria to assist in decision making under conditions of uncertainty. Commonly, this means deciding whether two treatments yield different results based on an experiment. The actual decision or interpretation must remain the responsibility of the researcher and, thus, can and should contain elements of subjectivity. There is, therefore, a need to separate the objective (statistical) analysis of data from its interpretation as well as to present all of the relevant information concerning the objective analysis. Current deficiencies in presentation include (i) failure to specify the assumptions on which an analysis is based, (ii) failure to provide sufficient quantitative information, (iii) failure to match the statistical analysis to the experimental situation, and (iv) failure to consider whether the assumptions on which an analysis has been based have been violated to an extent that would invalidate the results. Examples of these deficiencies from recent issues of this journal are cited with, where possible, suggestions on how the presentation could be improved. The present unfortunate situation is due to misinformation or lack of information, which stems in part, if not completely, from the ritualistic use of a few statistical methods and, in particular, statistical packages for computers that simply present the results of applying these methods, properly or otherwise.