Stability of Teacher Value-Added Rankings Across Measurement Model and Scaling Conditions

Value-added assessment methods have been criticized by researchers and policy makers for a number of reasons. One issue includes the sensitivity of model results across different outcome measures. This study examined the utility of incorporating multivariate latent variable approaches within a tradi...

Full description

Saved in:

Bibliographic Details
Published in:	Applied measurement in education 2017-07, Vol.30 (3), p.196-212
Main Authors:	Hawley, Leslie R., Bovaird, James A., Wu, ChaoRong
Format:	Article
Language:	eng
Subjects:	Achievement Tests Correlation Educational tests & measurements Elementary School Students Grade 1 Grade 2 Grade 3 Monte Carlo Methods Monte Carlo simulation Multivariate Analysis Reliability Scaling Standardized Tests Statistical Analysis Teacher Evaluation Value Added Models
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Value-added assessment methods have been criticized by researchers and policy makers for a number of reasons. One issue includes the sensitivity of model results across different outcome measures. This study examined the utility of incorporating multivariate latent variable approaches within a traditional value-added framework. We evaluated the stability of teacher rankings across univariate and multivariate measurement model structures and scaling metric combinations using a cumulative cross-classified mixed effect model. Our results showed multivariate models were more stable across modeling conditions than univariate approaches. These findings suggest there is potential utility in incorporating multiple measures with teacher evaluation systems, yet future research will need to evaluate the degree to which models recover known population parameters via Monte Carlo simulation.
ISSN:	0895-7347 1532-4818