ABSTRACT
Introductory statistics courses typically require responses which are approximately normal and independent of one another. This chapter introduces the methods that allow correlation to be taken into account. When modeling, correlation can be considered for normal and non-normal responses. The chapter focuses on recognizing data structures that may imply correlation, introducing new terminology for discussing correlated data and its effects. It considers the potential problems correlated outcomes may cause and why we need to take correlation into account when modeling. Correlated data is encountered in nearly every field. Correlated data often takes on a multilevel structure. The chapter presents two case studies on dams and pups, and tree growth. Correlated outcomes provide less information than independent outcomes, resulting in effective sample sizes that are less than the total number of observations. Correlation is likely and should be accounted for if basic observational units are aggregated in ways that would lead us to expect units within groups to be similar.
