Skip Navigation
Funding Opportunities | Search Funded Research Grants and Contracts

IES Grant

Title: Methods for Addressing Measurement Error Issues in Longitudinal Educational Studies
Center: NCER Year: 2016
Principal Investigator: Wang, Chun Awardee: University of Minnesota
Program: Statistical and Research Methodology in Education–Early Career      [Program Details]
Award Period: 2 years (9/1/20168/31/2018) Award Amount: $195,382
Goal: Methodological Innovation Award Number: R305D170042
Description:

Co-Principal Investigator: Gongjun Xu

The purpose of this project is to investigate different methods of using a two-stage framework to address measurement error in the estimation of latent theta scores obtained from standardized tests through item response theory (IRT). In the two-stage approach, an appropriate measurement model is first fitted to the data, and the resulting theta scores are used in subsequent analysis. Potential benefits of this approach include clearer definition of factors, convenience for secondary data analysis, convenience for model calibration and fit evaluation, and avoidance of improper solutions. Measurement errors are accounted for in the second-stage statistical analysis by combining the measurement error model and structural model in the same analysis or via a corrective approach, in which estimated thetas are first used in regression analysis but then the bias of the estimates of the regression coefficients and their standard errors are corrected in a post-hoc fashion.

The researchers will first complete the mathematical theory needed for three of the five proposed methods of accounting for measurement error in test scores. After programming everything in the R software package, the researchers will conduct simulation studies to evaluate the performance of the proposed approaches in a number of conditions similar to those found with national and state education-related datasets. The researchers will use the National Education Longitudinal Study of 1988 (NELS:88) science and math test data for illustration of the five methods of accounting for measurement error. By the end of the project, the researchers expect to release a user-friendly version of the software for running each of the methods, and will disseminate the results at conferences and in peer-reviewed journals.


Back