A Comparison of Multilevel Versus Standard Prediction Algorithms in the Presence of Nested Data

• In Event: Advances in Multilevel Modeling

Abstract

Many studies in educational research involve the collection and analysis of nested data, such as when data are collected from students who are situated within classrooms which themselves are nested within schools resulting in a three-level data structure. To predict categorical outcomes, there are a multitude of classification methods and recommendations are available. However, these algorithms were built specifically for non-nested data, and there has been very little research exploring the performance of classification algorithms on multilevel data. Our current study will use Monte Carlo simulation to compare the predictive performance among several multilevel prediction algorithms and several standard prediction algorithms under a variety of manipulated conditions that are inherent to multilevel data.

Authors

• Brennan Register, University of Maryland
  Presenting Author

• Tracy Sweet, University of Maryland
  Non-Presenting Author