Bayesian Data Analysis and New Opportunities for Ecologists

Abstract

Ecological systems are complex. Therefore, scientists rely on statistical tools to infer patterns and causation into their data. This has traditionally been done by employing a frequentist method, which defines probability as the relative frequency of the occurrence of a repeated event. The probability of observing similar or more extreme data than the collected dataset is then estimated given a null hypothesis. However, as new technology has emerged, more complex data have become available and conventional approaches to statistical inference have had limited success in capturing the complex structure of these datasets and therefore also in solving complicated modern issues such as climate change. A literature review shows that ecological research employing a Bayesian method of inference is gaining momentum. Importantly, in Bayesian statistics probability is defined as a degree of belief in the likelihood of an event to occur by incorporating uncertainty. Then, the probability of a hypothesis being true is evaluated conditional on available data and prior knowledge on parameters of interest. Furthermore, Bayesian inference allows to include not only uncertainty but also complex structures by allowing models to be built on multiple levels. Through a case study, the present work highlights the flexibility a Bayesian data analysis can provide, leading to more information rich results compared to conventional approaches. Additionally, the case study also underlines the importance of including multiple analytical approaches in order to reduce bias when determining the best possible explanation for the collected data and observed phenomena.