Abstract
While correlated data methods (like random effect models and generalized estimating equations) are commonly applied in practice, students may struggle with understanding the reasons that standard regression techniques fail if applied to correlated outcomes. To this end, this article presents an in-class activity using results from Monte Carlo simulations to introduce the impact of ignoring the correlation between outcomes by applying standard regression techniques. This activity is used at the beginning of a graduate course on statistical methods for analyzing correlated data taken by students with limited mathematical backgrounds. Students gain the intuition that analyzing correlated outcomes using methods for independent data produces invalid inference (i.e., confidence intervals and p-values) due to underestimated or overestimated standard errors of the effect estimates, even though the effect estimates themselves are still valid. While this standalone 90-minute in-class activity can be added at the beginning of an existing course on statistical methods for correlated data without any further changes, techniques for reinforcing students’ intuition throughout the course and applying this intuition to teach sample size and power calculations for correlated outcomes are also discussed. Supplementary materials for this article are available online.
Original language | English (US) |
---|---|
Pages (from-to) | 154-164 |
Number of pages | 11 |
Journal | Journal of Statistics and Data Science Education |
Volume | 30 |
Issue number | 2 |
DOIs | |
State | Published - 2022 |
Bibliographical note
Funding Information:I thank the editor, associate editor, and three anonymous reviewers for their useful suggestions that improved the clarity and content of this work.
Publisher Copyright:
© 2022 The Author(s). Published with license by Taylor & Francis Group, LLC.
Keywords
- Generalized estimating equations
- Mixed effect models
- Random effect models