ABSTRACT
Within the biological sciences, there is a significant shift taking place regarding the way data and other observations are being used to make inferences about biological phenomena and processes of interest. This shift is beginning to spill over into environmental toxicology and chemistry (Newman 2008). However, perhaps in no other field is this more apparent than in the wildlife sciences, where information theoretic approaches to model selection (Burnham and Anderson 2002; Johnson and Omland 2004) have become widely adopted. In other arenas, Bayesian approaches are being promoted as alternative frameworks that perhaps provide a more informative understanding of the data in hand (Clark 2005; Link and Albers 2007; Ellison 2004). However, these novel approaches are not merely alternative statistical “tests”; rather, they represent entirely different philosophical approaches, a fact that may not be entirely grasped by many practicing ecologists (Ellison 2004; Dennis 1996). Thus, to really understand the issues associated with the use of these novel approaches, one must investigate them in the context of the philosophy of science. In fact, it has been argued that much misunderstanding about statistics among graduate students arises not from a lack of technical knowledge about when and how to apply certain tests, but rather from a lack of training on “how to conduct science as an integrated process from hypothesis creation
7.1 Introduction .................................................................................................. 173 7.2 The Role of Models in the Scientific Method ............................................... 174 7.3 Statistical Models.......................................................................................... 177
7.3.1 Null Hypothesis Testing ................................................................... 177 7.3.2 Modeling Data .................................................................................. 179
7.3.2.1 The General Linear Model ................................................ 179 7.3.2.2 Generalized Linear Models ............................................... 182 7.3.2.3 Mixed-Effects Models ....................................................... 184
7.3.3 Information Theoretic Approaches................................................... 187 7.3.4 Bayesian Statistics ............................................................................ 189
7.4 Conclusions ................................................................................................... 193 References .............................................................................................................. 194
through statistical analysis” (Boyles et al. 2008). Similarly, I would argue that to effectively incorporate novel quantitative approaches into their own toolbox, practicing wildlife toxicologists need to be fully aware of how such approaches fit within their overall approaches to conducting science, how they relate to the actual biological hypotheses of interest, and how they can be reported in such a way as to promote the generation of new knowledge-not just “statistical ritual” (Guthery 2008).
