In this presentation I will discuss the major ways clustered data can be analyzed, especially when we only care about lower-level research questions, along with results from my own work showing how to avoid lower-level fixed effect coefficient (slope) bias, and why it's especially important for multilevel modeling alternatives. If time permits, I'll also present some work on statistically comparing two predictors' slopes from the same model, including for frequentist and Bayesian multilevel models. R code and sample data will be provided.
Dr. Liz Sanders is Associate Professor in the College of Education's Measurement & Statistics program. Her research focuses on evaluating applied analytic methods for handling clustered data. Outside of work, Liz loves running.
What if I only care about L1 X-Y relations? Clustered data analytic options