A Sequential Generalized Nonparametric Classification Method for Small-Scale Cognitive Diagnostic Assessment

Area of Science:

• Educational Measurement
• Cognitive Psychology
• Psychometrics

Background:

• Small-scale classroom assessment is crucial but lacks suitable cognitive diagnostic tools.
• Polytomous items offer rich skill measurement but pose estimation challenges for existing models.
• Parametric models need large samples; nonparametric models typically handle only dichotomous data.

Purpose of the Study:

• To develop a nonparametric cognitive diagnosis method for polytomous data suitable for small-scale assessments.
• To address the gap between practical classroom needs and advanced diagnostic modeling capabilities.

Main Methods:

• Proposed the seq-GNPED method, extending generalized nonparametric classification to polytomous responses.
• Introduced weighted ideal category response and a collapsed class iterative algorithm.
• Utilized simulations and empirical data for validation.

Main Results:

• seq-GNPED demonstrated robust and accurate diagnostic performance in small sample conditions.
• The method effectively handles polytomous response data where parametric models fail.
• Achieved fine-grained cognitive diagnosis tailored for classroom settings.

Conclusions:

• seq-GNPED provides a practical, nonparametric solution for cognitive diagnosis with polytomous items in small samples.
• This method bridges the gap between theoretical models and real-world classroom assessment needs.
• Enables more precise measurement of student skills and cognitive processes in typical educational environments.