Chapter 6: Integrating Statistical Methods and Results into Your Writing
James R. Bence and Daniel B. Hayes
Over the past 30 years, the number and diversity of statistical analyses in fishery publications have undergone a nearly exponential increase. This increase reflects greater pluralism in inference paradigms (i.e., other than hypothesis testing) and a greater variety of methods and models (Figure 1). These changes are similar to changes that are occurring in other ecological and resource management journals (Hobbs and Hilborn 2006; Sleep et al. 2007). We believe that this change largely reflects the increase in available computing power, which continued to essentially double every 2 years over this period (e.g., Twist 2005). This increase in computing power has made increasingly complex and computationally challenging analyses practical as software programs that support such analyses have become available.
Although the ability to easily accomplish more appropriate and powerful analyses than had been possible is beneficial, the increase in the diversity of statistical results being incorporated into manuscripts poses a substantial challenge for authors and reviewers. Authors need to understand and describe increasingly complex procedures that often are tailored to the problem at hand. Reviewers and editors need to consider whether the approaches are appropriate and sufficiently well described and whether results are presented clearly and adequately. In the early 1980s, knowledge of linear regression, t-tests, and simple analysis of variance (ANOVA) models encompassed the statistical background needed for writing, reviewing, and interpreting most fishery publications. Currently, an analyst may need to know resampling methods, nonlinear model fitting, maximum likelihood for nonnormal distributions, models mixing fixed and random effects, information-theoretic model selection procedures, and much more. This larger “statistical toolbox” has led to an increased need and demand for training in statistical methods, inference paradigms, and their role in science, both through formal university offerings and through specialized short courses for working professionals (e.g., Hobbs and Hilborn 2006; Butcher et al. 2007; del Rio et al. 2007; Boyles et al. 2008). The expansion of the tools used to make inference about the population of interest also has greatly increased the challenge of concisely but clearly and accurately communicating what analyses were performed and what results were obtained.