A Sequential Response Model for Analyzing Process Data on Technology-Based Problem-Solving Tasks.

Area of Science:

• Educational technology
• Psychometrics
• Data science

Background:

• Student responses in technology-based problem-solving tasks can be modeled as a discrete time stochastic process.
• The conditional Markov property suggests that future responses depend only on the current state, given student abilities.
• Accurate inference of problem-solving ability is crucial for effective educational assessment.

Purpose of the Study:

• To propose a novel sequential response model (SRM) for inferring problem-solving ability.
• To incorporate comprehensive information from the response process for more effective ability estimation.
• To utilize a Bayesian approach for robust parameter estimation within the SRM.

Main Methods:

• Developed a sequential response model (SRM) based on a discrete time stochastic process with a conditional Markov property.
• Employed a Bayesian approach for parameter estimation, integrating response sequence data.
• Validated the model using a Monte Carlo simulation study and an illustrative example with PISA 2012 data.

Main Results:

• Monte Carlo simulations demonstrated that the model parameters were well-recovered.
• The SRM effectively incorporates comprehensive information from student response sequences.
• The illustrated example showed additional gains in understanding the response process using the proposed model.

Conclusions:

• The proposed sequential response model (SRM) offers a more effective method for inferring problem-solving ability.
• The Bayesian approach within the SRM enhances parameter estimation by utilizing rich response data.
• The model provides valuable insights into the student response process in interactive assessments.