You are here:
Analyzing mixing systems using a new generation of Bayesian tracer mixing models
Citation:
Stock, B., A. Jackson, E. Ward, A. Parnell, D. Phillips, AND B. Semmens. Analyzing mixing systems using a new generation of Bayesian tracer mixing models. PeerJ. PeerJ Inc., Corte Madera, CA, , e5096, (2018). https://doi.org/10.7717/peerj.5096
Impact/Purpose:
Scientists from EPA’s Western Ecology Division, NOAA National Marine Fisheries Service, and universities in the U.S., Canada, Ireland, and the U.K. have submitted a manuscript to Methods in Ecology and Evolution describing a new stable isotope mixing model called MixSIAR. Some biologically important elements such as carbon, nitrogen, oxygen, and hydrogen occur in several stable (non-radioactive) isotopic forms. Physical, chemical, and biological processes cause variation in the isotopic ratios of various chemical and biological materials which can be used as a sort of chemical signature to trace their movement through the environment. When these isotopic ratios are measured in both mixtures and the sources contributing to them, mixing models can be used to estimate the mixing proportions of the sources. Environmental applications include determining pollutant sources in polluted air or water, and quantifying ecological food webs (who eats who). Building on simpler models developed at EPA WED in the early 2000’s, the authors have devised a comprehensive, flexible model based on Bayesian statistics that can incorporate a wide variety of features in mixing analyses as needed. These include higher limits on the number of sources, accounting for sample variability, dependence on chemical concentrations, use of prior information on likely source proportions, inclusion of covariates that may affect the proportion estimates, and a number of others. This new model promises to allow one-stop-shopping for a wide variety of environmental mixing analyses and designs.
Description:
The ongoing evolution of mixing models has resulted in a confusing array of software tools that differ in terms of data inputs, model assumptions, and associated analytic products. Here we introduce MixSIAR, an inclusive, rich, and flexible open-source Bayesian stable isotope mixing modeling framework with associated software. MixSIAR allows users to easily perform most state-of-the-art calculations currently available in existing tools, while also providing more advanced statistical parameterizations (e.g. fixed effects and hierarchical random effects of covariates) that previously were only available through custom model development. Using MixSIAR as a foundation, we provide guidance for the implementation of mixing model analyses. We begin by outlining the practical differences between the mixture data error structure formulations available in MixSIAR, and relate these error structures to common mixing model study designs in ecology. We also discuss the options available for source data inputs (raw data versus summary statistics), and provide guidance for combining sources. Because MixSIAR has the option to specify informative priors on source proportion contributions, we outline methods for establishing prior distributions and discuss the influence of prior specification on model outputs. Finally, we discuss the application of fixed and random effects in mixing models. Through MixSIAR, we have consolidated the disparate array mixing model tools into a single platform, diversified the set of available parameterizations, and provided developers a platform upon which to continue improving mixing model analyses in the future.
URLs/Downloads:
DOI: Analyzing mixing systems using a new generation of Bayesian tracer mixing models