ADVANCES IN STATISTICAL PHILOSOPHY AND EXPERIMENTAL DESIGN IN BEHAVIOURAL ECOLOGY: A NEW ABC FRAMEWORK ISBE 2008 post-conference symposium, Cornell University, Ithaca, New York, USA, August 15, 2008
We are planning to apply for possibility to organize a post-congress symposium for the next ISBE International Behavioral Ecology Congress in 2008 at Cornell University. We would like to solicit an active discussion, for which the framework of conference would provide us, behavioural ecologists with a wonderful opportunity. Our symposium would deal with recent advancements in the interpretations and analyses of behavioural and ecological data (please, find our proposed abstract below for more details). To achieve this, we would like to invite participants who would be interested in sharing their experience with a broad audience on the proposed statistical issues. If you are registered (or planning to register) for the main program of the conference, and are potentially willing to give a short talk in our proposed a post-congress symposium, please contact either László or Shinichi for more information: László Zsolt Garamszegi [EMAIL PROTECTED] Shinichi Nakagawa [EMAIL PROTECTED] PROPOSED ABSTRACT Advances in statistical philosophy and experimental design in behavioural ecology: a new ABC framework László Zsolt Garamszegi & Shinichi Nakagawa Department of Biology, University of Antwerp, Wilrijk, Belgium Department of Animal & Plant Sciences, University of Sheffield, Sheffield, UK We aim to bring together researchers with different interests in behavioural ecology to discuss recent significant developments in the interpretation of behavioural and ecological data. Analytical tools that incorporate statistical philosophies not relying on statistical significance have been highlighted in recent years, but are still in limited use in our field, because of common (mis)beliefs and familiarity with classical approaches. We suggest that conventional hypothesis testing based on statistical significance is often misleading or even incorrect in some circumstances, because it assesses biological hypotheses using an all-or-nothing criterion, instead of along a continuum. It also leads to publication bias, due to unreported non-significant effects, hindering the field from advancing as cumulative science. We propose an ABC framework, in which such problems can be powerfully treated. The ABC framework will open up new horizons for testing evolutionary questions while it sets straightforward guidelines for researchers in the field to follow. In this framework, A stands for AIC (Akaikes Information Criterion), a popular tool of information theoretic IT (information theoretic) approaches that address the trade-off between model complexity and goodness of fit. This allows an objective assessment of the potentially large number of competing models that could describe the data equally well. B refers to Bayesian inference, in which new empirical evidence is combined with past knowledge to update or newly infer the probability of the hypotheses under test. Bayesian approaches also harness us with capabilities of parameter estimation in a way that alternative to the classical framework. C emphasizes the importance of confidence intervals, which give the precision of parameter or effect size estimates. Effect size estimates and their confidence intervals should be at the heart of statistical inference because they relate to biological importance in a way that statistical significance does not. Effect size estimates also facilitate meta-analysis, which has recently established itself as an essential tool for quantitative review in the field. The three components of our ABC framework complement each other. We would like to focus on the basic philosophy behind these approaches by considering pro- and contra arguments, and provide a broad array of examples of biological questions that can be tested in correlative or experimental designs, in intra- or inter-specific contexts, and in different taxonomic groups.
