WK 50 - Mixed Effects Modeling Primer: Selecting Error Distributions, Accounting for Correlation, and Building Statistically Sound Mixed Effects Models for Ecological Data

Session Description: Building robust and statistically sound ecological mixed effects models is challenging. Sample sizes and study lengths are often limited by funding and are often not as large nor as long as managers or researchers would like. Mixed effects models provide flexible tools to analyze ecological data, but a poorly-specified statistical model or violation of model assumptions can result in erroneous inference. The goal of the workshop is to provide guidance and considerations for ecologists building mixed effects models. This workshop will review how to apply mixed effects models, handle dependencies, and assess model fit. First, we will review error distributions for mixed effects models. Then, we will discuss how and when to account for spatial and temporal autocorrelation, correlated predictors, and additional dependencies that arise due to study design or sampling structure. Finally, we will review how to assess model fit using hypothesis tests and graphical tools. The workshop will provide hands on demonstration of how to handle these complex issues in programming language R. Participants will leave the workshop with experience in applying mixed effects models to ecological data and assessing assumptions and model fit. Using example datasets, this interactive workshop provides hands-on experience modeling different types of data and evaluating model assumptions and fit. Participants should use laptops with an installed copy of R/RStudio (https://posit.co/downloads/) and the glmmTMB and DHARMa packages.