A Markov Estimation Strategy for Longitudinal Learning Diagnosis: Providing Timely Diagnostic Feedback.

Area of Science:

• Educational Measurement
• Psychometrics
• Learning Analytics

Background:

• Timely diagnostic feedback is crucial for effective learning and teaching adjustments.
• Current longitudinal learning diagnosis models often lack timely feedback due to simultaneity estimation.
• The need for "assessment for learning" and formative assessment drives the search for improved diagnostic strategies.

Purpose of the Study:

• To propose and evaluate a novel Markov estimation strategy for longitudinal learning diagnosis.
• To compare the performance of the Markov strategy against simultaneity, anchor-item, and separated estimation strategies.
• To determine if the proposed strategy can provide timely diagnostic feedback.

Main Methods:

• A simulation study was designed to compare four estimation strategies.
• The strategies evaluated were simultaneity, Markov, anchor-item, and separated estimation.
• Performance was assessed based on parameter estimation and timeliness of feedback.

Main Results:

• The performance of the four strategies was highly consistent.
• The relative performance order was: simultaneity > Markov > anchor-item ≥ separated.
• The proposed Markov strategy offers a slight trade-off in parameter estimation accuracy for improved timeliness.

Conclusions:

• The Markov estimation strategy provides timely diagnostic feedback, supporting formative assessment.
• This strategy addresses the limitations of simultaneity estimation in current longitudinal models.
• The findings support the integration of timely feedback in educational assessment for improved learning and teaching.