A Tutorial on Centering in Cross-Sectional Two-Level Models |
| |
| Authors: | Nicholas D Myers Ahnalee M Brincks Mark R Beauchamp |
| |
| Institution: | 1. Department of Educational and Psychological Studies , University of Miami , Coral Gables , Florida nmyers@miami.edu;3. Department of Epidemiology and Public Health , University of Miami , Coral Gables , Florida;4. The University of British Columbia, School of Human Kinetics , Vancouver , British Columbia , Canada |
| |
| Abstract: | The primary purpose of this tutorial is to succinctly review some options for, and consequences of, centering Level 1 predictors in commonly applied cross-sectional two-level models. It is geared toward both practitioners and researchers. A general understanding of multilevel modeling is necessary prior to understanding the subtleties of centering decisions. A review of some high-quality journals within the broad discipline of exercise science provides evidence that multilevel modeling is used relatively infrequently in this field. Therefore, a secondary purpose is to introduce Measurement in Physical Education and Exercise Science readers to some core facets of multilevel modeling within the framework of this tutorial. A relevant dataset is used to demonstrate potential consequences of different centering decisions within a multilevel model. Depending on the model and the data, different centering decisions can exert non-trivial influence on the meaning of some model parameters, results from fitting the model, and subsequent conclusions. |
| |
| Keywords: | grand mean centering group mean centering raw score centering within cluster multilevel modeling hierarchical linear modeling |
|
|
