By Julie Garai (The University of the South); Marina Ptukhina (Whitman College); Walt Stroup (University of Nebraska–Lincoln)
Students in an intermediate statistics course gain skills to analyze data independently. Such an experience without an introduction to study design leaves students in danger of developing models that are misleading or downright incorrect. To help students gain a better understanding of how study design relates to building a model, we propose adopting a method introduced in Stroup (2013), What Would Fisher Do (WWFD). This method includes two preliminary steps that when combined build a traditional ANOVA table. Students are taught to distinguish between how the units of observation are organized (Design Structure: Step 1), how the treatments are applied (Treatment Structure: Step 2), and how to combine the structures into an ANOVA table (Step 3). Associated degrees of freedom are considered at all steps. With WWFD, students learn a method of organizing study components to properly account for extraneous variability and thus to model data as it naturally arises.