Statistical methods are used by ecologists to: (1) identify meaningful patterns in their data and (2) model the observed patterns for the purpose of making predictions. These two tasks are often related, but in this article the focus is on the latter, which is the domain of 'modelbased statistics'. The former, including the formal procedures of experimental design and hypothesis testing, can be considered the domain of 'design-based statistics'. Still, it is important to remember that model-based predictions serve a variety of purposes, including exploring the implications of multiple hypotheses, anticipating the outcomes of possible experimental designs, and providing support for ecosystem management decisions.
To present the range of methods available for statistical ecological modeling, this article proceeds from the most restricted and familiar setting: linear regression analysis, to increasingly general techniques which sequentially drop some limiting assumptions. This will allow the reader to see the relation among various methods and to proceed through the presentation, building on his or her ability and experience. Of course, in this relatively brief overview, there will not be space for detailed explanations and examples. However, the goal is to provide a structured framework for understanding the methods and to rely on the sources listed for further reading to provide the details.
Was this article helpful?