Comparing Multilevel Models and the Averaging Method for Estimating Teacher Ratings

• In Event: Sheraton Roundtable Session Seventeen
  In Roundtable Session: Data-Driven Instructional Design (Table 2)

Abstract

The purpose of this study was to examine the usage of multilevel measurement models (MLM) for estimating teacher evaluation ratings, compared to the traditional averaging method. Through a preliminary simulation, we show that MLMs produce larger mean estimates when a binary predictor (demographics information, gender for example) is included at the student level, compared to the averaging method. Contrarily, MLMs produce smaller mean estimates when a predictor is added at the teacher level. MLMs are also relatively robust to biases introduced at both levels. These simulation results will serve as a guide for the analysis of real data using these methods. Data collection will begin this coming fall.

Authors

• Guangyun Liu, University of Iowa
  Presenting Author

• Xi Wang, University of Iowa
  Presenting Author

• Jonathan Templin, University of Iowa
  Non-Presenting Author