ABI Development: An extensible software platform for integrating multiple sources of data and uncertainty using hierarchical statistical models

点击翻译按钮获取中文摘要

Hierarchical statistical models allow estimation of patterns in complex biological data while accounting for relationships such as temporal or spatial patterns or shared sampling units. A great variety of analysis algorithms for hierarchical models have been developed by statistical researchers but are unavailable to practitioners such as experimental or field biologists. These include many types of Markov chain Monte Carlo, as well as sequential Monte Carlo, importance sampling, approximate Bayesian computation, and other numerical methods and approximations. In addition, there are many higher-level algorithms that use these as components of methods for model selection, model averaging, maximum likelihood estimation, generating predictions, and more. This project will involve development of an open source, extensible software environment for flexible composition of hierarchical models and algorithms. The software will include low-level components in which algorithms will be executed for speed, high-level components in which algorithms can be composed and managed from the R statistical software environment, and middle-level components to interface the first two. Many algorithms will be implemented and disseminated for application using the new software. Moreover, it will provide a foundation for ongoing development and sharing of new and improved algorithms in the future.Hierarchical statistical models are used in many domains of biology to provide robust conclusions and management guidance that harness all available data. Areas of application include wildlife conservation and management, ecosystem processes such as carbon cycling, organismal growth and development, and cellular biochemical networks. In all of these areas, biologists need to use complicated data to estimate the processes and rates of change occurring in their study system. This project will provide a next generation of software to make available numerous algorithms to many researchers to achieve this goal. These algorithms will facilitate research workflows by allowing researchers to extract the most information from their data in an efficient manner.