|
|
|
| Non-parametric Methods for Longitudinal Cognitive Diagnosis |
| ZHENG Tian-peng1，GUO Lei2,3，BIAN Yu-fang1,4,5 |
| 1. Collaborative Innovation Center of Assessment toward Basic Education Quality (CICA-BEQ) at Beijing Normal University, Beijing 100088, China; 2. Faculty of Psychology, Southwest University, Chongqing 400715, China; 3. Southwest University Branch, Collaborative Innovation Center of Assessment toward Basic Education Quality, Chongqing 400715, China; 4. Child and Family Education Research Center at Beijing Normal University, Beijing 100088, China; 5. Institute of Mental Health and Education at Beijing Normal University, Beijing 100088, China |
|
|
|
|
Abstract This research proposed four concise longitudinal non-parametric diagnostic methods: longitudinal non-parametric classification (LNPC), longitudinal weighted non-parametric classification (LWNPC), longitudinal generalized non-parametric classification (LGNPC), and longitudinal weighted generalized non-parametric classification (LWGNPC), by leveraging the connection of students’ knowledge state and the ideal response pattern between adjacent time points. The performance of the four new methods was evaluated by two simulation studies and an empirical study. The results of simulation studies showed that: (1) the established link can improve the accuracy of longitudinal classification; (2) compared to the HO-HMM model, the new methods can achieve similar precision while being less affected by the sample size; (3) the new methods outperformed the Long-HDD method in terms of accuracy, and the LNPC and LWNPC still performed well even with low-quality items. The results of the empirical study showed that the four new methods were highly consistent with the HO-HMM model and Long-HDD classification. We recommend using the LWNPC method to analyze the real data.
|
|
Published: 30 April 2023
|
|
|
|
|
|
|
