Using Multidimensional Latent Regression to Link Between Large-Scale Educational Survey Assessments

• Activity Number: 302
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Social Statistics Section
• Author(s): Yue Jia*+ and Xueli Xu
• Companies: Educational Testing Service and Educational Testing Service
• Keywords: Latent Regression ; Linking/Prediction ; Large scale educational survey assessments

Abstract

Equating and linking is a common practice in educational programs. A number of statistical approaches have been applied in linking within educational programs and between different programs. However, little is known about what a general approach would be in linking educational survey assessments due to their uniqueness in inferring proficiency estimates. This study will focus on a latent regression approach and its application in linking two large scale survey assessments.

Large scale educational survey assessments, such as the National Assessment of Educational Progress (NAEP) and the Trends in International Mathematics and Science Study (TIMSS),commonly use a combination of Item Response Theory models and latent regression population models to estimate distributions of underlying proficiencies for student groups of interest. To link between different survey assessments of such, a regression procedure was developed to model the relationship between proficiencies yielded from these two assessments. This model was then used to project scores from one assessment onto the scale of the other. The procedure was applied in linking NAEP to TIMSS in their 2011 administration.