Study design and sampling decisions in mixed-effect model analyses

Study design and sampling decisions in mixed-effect model analyses

  • Beginn: 01.12.2015
  • Ende: 03.12.2015
  • Vortragende(r): Prof. Dr. Niels Dingemanse
  • The brain and eight arms of SQuID are: Hassan Allegue + Denis Réale (Université du Québec à Montréal, Canada), Yimen Araya-Ajoy + Niels Dingemanse (MPI for Ornithology), Ned Dochtermann (University of North Dakota, USA), Laszlo Garamszegi (Estación Biológica de Doñana, Spain), Shinichi Nakagawa (University of Otago, New Zealand), Holger Schielzeth (University of Bielefeld, Germany), and Dave Westneat (University of Kentucky, U.S.A.).
  • Ort: Seewiesen
  • Raum: Seminar Room House 5
  • Gastgeber: IMPRS for Organismal Biology
  • Kontakt: mhieber@orn.mpg.de
Study design and sampling decisions in mixed-effect model analyses
This unique three-day course is given by the members of SQuID, which stands for Statistical Quantification of Individual Differences. The SQuID-group consists of a group of behavioural ecologists interested in teaching how one may empirically study variation in labile phenotypic traits (i.e. traits that have a hierarchical structure). For example, behaviour and physiology vary both among populations, among individuals within populations, and within individuals. A number of major challenges in the analysis of labile phenotypic traits will be addressed in this course. First, we detail the utility of applying a reaction norm framework, and outline how this may be achieved by estimating ‘phenotypic equations’ using mixed-effect models. Second, we discuss optimal study designs and sampling decisions, and detail how those might affect statistical power, accuracy and precision of estimated parameters. Finally, the SQuID-group has developed an educational package that allows users i) to learn to understand and use mixed-effect models, and ii) simulate and visualize hypothetical datasets from which data can be sampled. This package will therefore be used to iii) investigate how study design and sampling decisions affect parameter estimation. We will focus on population-average norms of reaction, among-individual variation in reaction norm intercepts (repeatability) and slopes (plasticity), and how unknown (temporally auto-correlated) patterns in the environment can bias such parameters in sampled data. The course exists of 6 half-day blocks, taught by (a) member(s) of the SQuID-group, where the last blocks will be reserved to applying the learnt concepts to own data.

The brain and eight arms of SQuID are: Hassan Allegue, Yimen Araya-Ajoy, Niels Dingemanse, Ned Dochtermann, Laszlo Garamszegi Shinichi Nakagawa, Denis Réale, Holger Schielzeth, and Dave Westneat.

Zur Redakteursansicht