Advances in tracking technology in the past decades strongly enhanced our ability to track animal movements, and generated an ever increasing body of literature pertaining which gave birth to the field of Movement Ecology. The level of statistical analysis in ecology journals is considerable higher nowadays, developing and applying numerous models for animal movement data, particularly to identify behavioral patterns through movement; e.g. independent mixture models, state-space models and correlated random walk models to name but a few. Movement models are usually introduced to the community as reproducible (i.e. with the same data, shared in the paper, the same results should be expected) and generalizable (i.e. the method could be used for similar questions but different data). With a variety of models to study animal movement, we revisited the most popular methods used to identify behavioral patterns in animal movement. We selected frequently cited papers applying these methods, for which the data was publicly available. We then assessed the correctness of the statistical models (e.g. the data was in the right format, the assumptions were verified and respected), the reproducibility of the results, and ultimately the appropriateness of the method for the ecological questions tackled in these studies.
Results/Conclusions
For this poster, we present results from seven selected methods applied to a range of species. Half of the papers investigated did not make use of the most suitable method to analyze their data. From the remaining ones, we were only able to reproduce the results twice out of 4 papers. Moreover, we found that model assumptions had often been violated, specifically with regard to the temporal regularity of the movement data, a key statistical assumption for movement. We thus re-analyzed their data using a more suitable method and wrote up recommendations for these specific cases. Finally, to illustrate how to choose the appropriate method based on the ecological questions and structure of the data, we present a complete example applied to novel seabird movement data to better understand their movement decisions in a changing meteorological environment. Statistical errors are common across a wide range of disciplines and even in the most prestigious journals. Ecological research is not immune from statistical misconceptions which can be dangerous as they may lead to inaccurate conclusions. We hope that this work will be a small step towards statistically sound science in the our field of Movement Ecology.