Assessment of five statistical approaches in finding out the over- and under-represented plant species in traditional medicines in Nepal

Background: In quantitative ethnobotany, different statistical tools and approaches have been used to compare medicinal floras with the overall flora of a given area and to investigate over- and under-represented medicinal plant families. Methods: In this study, we analyzed a dataset of medicinal plant species of Nepal to evaluate their usefulness in traditional medicines in Nepal. We compared five different statistical methods (Bayesian, Binomial, Least-square, Log-transformed, and Negative-binomial) to test which plant families are more likely to harbor more medicinal plant species that are being used. Results: Among the tested methods, least-square method was found more pertinent in the sense that the over-represented medicinal plant families of this approach and that being used in traditional medicines resembled the greater affinity. For small and community level datasets, negative-binomial analysis was found pertinent. Thus, it is particularly important to combine statistical approaches for small, and moderate-sized data to avoid inherent methodological biases. The combined approach of all five statistical methods generated the over-represented plant families in the following order Moraceae, Cucurbitaceae, Zingiberaceae, Rutaceae, Solanaceae, Euphorbiaceae, Lamiaceae, and Anacardiaceae, which is close to the result of 5% significance level of negative-binomial analysis. Conclusions: This analysis allowed us to identify plant families which are apparently underutilized and which apparently are substantial sources of traditional medicines in Nepal. This study also contributed to the discussion of the methodology to test the non-random theory of medicinal plant selection. Similar assessments are needed for finding out the best methods for identifying the over- and under-utilized plant species. Keywords: Statistical methods, regression, binomial, Bayesian, medicinal plants, over-represent, under-represent, Nepal.