Evaluation of Student Performance through a Multidimensional Finite Mixture IRT Model

Area of Science:

• Educational Measurement
• Psychometrics
• Academic Analytics

Background:

• In the Italian academic system, students can either take exams immediately or postpone them, leading to missing data for postponed exams.
• Traditional student performance evaluations often overlook non-attempted exams, potentially skewing results.

Purpose of the Study:

• To develop a novel approach for evaluating student performance that incorporates non-attempted exams.
• To provide a more comprehensive and accurate assessment of a student's academic journey.

Main Methods:

• Utilized an item response theory (IRT) model with two discrete latent variables: student performance and exam selection priority.
• Employed a within-item multidimensionality approach to handle non-ignorable missing observations, where attempted exams inform performance measurement.
• Incorporated individual covariates into the structural part of the model.

Main Results:

• The proposed IRT model effectively evaluates student performance by accounting for both attempted and non-attempted exams.
• The model demonstrates that non-attempted exams provide valuable information for performance assessment.
• The inclusion of covariates allows for the analysis of factors influencing student performance and exam selection.

Conclusions:

• The developed IRT model offers a robust framework for assessing student academic performance in systems with flexible exam scheduling.
• This approach provides a more nuanced understanding of student progress by integrating data from non-attempted exams.
• The methodology has implications for educational policy and student support strategies.