Analyzing process data alongside traditional item responses to obtain more accurate imputation, offering a deeper understanding of respondent behavior and enhancing the quality of imputing missing responses

This project aims to enhance proficiency estimation in large-scale assessments by improving missing-data imputation techniques. Specifically, the study focuses on refining Multiple Imputation with Denoising Autoencoders (MIDAS)—a deep learning-based approach—by incorporating item response time as an additional contextual feature. Unlike traditional item response theory (IRT) or regression-based methods, which rely on strong assumptions, the proposed approach leverages the flexibility of deep learning to better handle the complexity and dimensionality of large-scale assessment data.

Faculty Supervisor:

Ying Cui

Student:

Partner:

ETS Canada;ETS Global;Educational Testing Service

Discipline:

Sociology

Sector:

Education; Professional, scientific and technical services

University:

University of Alberta

Program:

Accelerate

Current openings

Find the perfect opportunity to put your academic skills and knowledge into practice!

Government

International

Load More

International

Load More