Application of Bayesian inference for estimating parameters of probability distributions of concentration ratios

Concentration ratios (CRs) are used to derive activity concentrations in wild plants and animals. Usually, compilations of CR values encompass a wide range of element-organism combinations, extracted from different studies with statistical information reported at varying degrees of detail. To produce a more robust estimation of distribution parameters, data from different studies are normally pooled using classical statistical methods. However, there is inherent subjectivity involved in pooling CR data in the sense that there is a tacit assumption that the CRs under any arbitrarily defined biota category belong to the same population. Here, Bayesian inference has been introduced as an alternative way of making estimates of distribution parameters of CRs. This approach, in contrast to classical methods, is more flexible and also allows us to define the various assumptions required, when combining data, in a more explicit manner. Taking selected data from the recently compiled wildlife transfer database (http://www.wildlifetransferdatabase.org/) as a working example, attempts are made to refine the pooling approaches previously used and to consider situations when empirical data are limited.