Numerical analysis and group formation in palynology

Abstract
All areas of paleontology are experiencing a rapid rise in the use of statistical treatments of data, and palynology is no exception. The range of these treatments open to palynologists is great; however all involve (1) putting the data into a format wherein it can be analyzed statistically, (2) performing the statistical test, and (3) transforming the numerical results of the test into a format from which interpretation can be made. It should be possible to choose approaches at all three of these stages which maximize the information which can be obtained, while taking into consideration the special problems inherent in palynology, such as the extreme transportability of palynomorphs and the differential production of pollen and spores by different species of plants. This study compares several different statistical approaches to a specific palynological problem drawn from the Permian of Oklahoma, pointing out the advantages and disadvantages of the various methods. It suggests that the most meaningful and general method of forming recurrent groups with stratigraphic or paleoecological significance, is through the use of a nonparametric rank statistic followed by a ramifying linkage cluster analysis, although statistical tests based on counts such as product‐moment correlation and interpretive techniques such as factor analysis may be very useful under certain circumstances.