Sample Size Dependence in Measures of Proportional Similarity

Publisher Website
Google Scholar
Abstract
The 2 commonly presented measures of proportional similarity differ in that one (PSI) is independent of the relative sizes of the samples being compared, but the other (PSD) is not. Using hypothetical data sets in which the proportions of entries in each sample remain constant but the ratio of total sample sizes is varied, the differences between these measures are characterized. Both give the same value when the 2 sample sizes are equal. When the ratio of sample sizes .noteq. 1, the value of PSD is usually less than the value of PSI. Since PSD is often affected more by sample size ratio than by proportionate compositional similarity (it may be a function of sample size ratio only), the use of PSI is suggested, except when it is intended that the index reflect sample size differences.