Date
7-9 March 2022
09:00  - 15:00 (Germany) / 16:00 - 22:00 (China); including a break for lunch/dinner

Location
online

Credit points
1 CP

Target Group
Doctoral Researchers (2nd cohort)

Contents

Day 1:
Linear models and linear mixed effects models in R (Part 1)
Lecturer: Uli Brose

Day 2:
Introduction to Structural Equation Modeling (SEM)
Lecturer: Shaopeng Wang

Linear models and linear mixed effects models in R (Part 2)
Lecturer: Uli Brose

Day 3
09:00  - 14:00 (Germany) / 16:00 - 21:00 (China)
Lecturers: John Connolly, Caroline Brophy, Rafael De Andrade Moral

Introduction to Generalised Diversity-Interactions (GDI) modelling to explain ecosystem functioning. Discussion of quantifying diversity as an explanatory variable in regression models. This includes an introduction to the R package DImodels for fitting Diversity-Interactions models (https://cran.r-project.org/web/packages/DImodels/vignettes/DImodels_Vignette.html).

Note on lectures

My two lectures will be somewhere between lectures and seminars discussing new ideas. I will be discussing GDI models and Caroline will be showing various analyses of such models using the DImodels package.

In addition to briefly describing DI models I want to raise various issues about the design and analysis of BEF experiments that I believe are very relevant to the efficiency of the analysis. Here I use efficiency in the technical statistical sense of the size of the SE attaching to key parameters in the models. I will discuss the following issues, using data from the BEF-China pilot study (Schmid et al, 2017) as appropriate:

• Design of BEF experiments including the discussion of design variable and proposining a new set of such variables that in some circumstances should increase analysis efficiency
• The inclusion of species identity in models
• The coverage of design space as affected by the design variables used
• Conditions for the validity of substitutive experiments and the inclusion of community density as defined by criteria other than numbers of species and individuals
• Inclusion of non-design variables in models
• Discussion of analysis of studies spanning several years at a significant time distance from the establishment of the experiment and the inclusion of additional experimental treatments in such studies.

I expect that some of the ideas may be controversial and I very much welcome discussion on those. It just seems to me at times that a lot of valuable information is not being explored in an analysis that cleaves to a strict protocol while at the same time ignoring issues that could seriously undermine some of the analysis.

Goal

· Obtaining the ability to run simple and mixed linear models to analyze the relationship between biodiversity and ecosystem functioning
· Knowing the strengths and limitations of SEM and how to use SEM to address own research questions
· Understanding how to use Generalized Diversity-Interactions models to address effects of species diversity, evenness and composition on ecosystem functioning.

Didactic Elements
Lectures, Hands-on training

You will need
Basic knowledge of statistics and linear models
Basic knowledge of R
Own laptop with R and R studio installed

Expected performance
Active participation, independent analysis of the course data sets, writing own R code, short presentation of own results.