Department of Teaching, Learning and Teacher Education
Date of this Version
Spring 10-17-2020
Document Type
Article
Citation
The Nebraska Educator, Volume 5 (2020), pp.92-104.
ISSN 2375-6853
doi: 10.32873/unl.dc.ne005
Abstract
Physical, behavioral and psychological research questions often relate to hierarchical data systems. Examples of hierarchical data systems include repeated measures of students nested within classrooms, nested within schools and employees nested within supervisors, nested within organizations. Applied researchers studying hierarchical data structures should have an estimate of the intraclass correlation coefficient (ICC) for every nested level in their analyses because ignoring even relatively small amounts of interdependence is known to inflate Type I error rate in single-level models. Traditionally, researchers rely upon the ICC as a point estimate of the amount of interdependency in their data. Recent methods utilizing an interval estimation of the amount of interdependency based the proportion of second-level variance between groups have been developed that avoid relying solely upon point estimates. The likelihood of committing a Type I error when using the interval estimation of the proportion of second-level variance remains unknown. The current project addressed this deficiency in knowledge utilizing simulated data to assess the accuracy of a 95% confidence interval estimation of the proportion of second-level variance (CI-PSLV). Standard errors tended to decrease as sample size increased, and the CI-PSLV captured the second level ICC in 95% of replications.
Included in
Applied Statistics Commons, Statistical Models Commons, Teacher Education and Professional Development Commons
Comments
© 2020 University of Nebraska