The code for SLR-B and eLasso methods with Q-matrix validation

The Q-matrix defines the associative structure between test items and latent attributes in cognitive diagnosis, and its mis-specification can lead to diagnostic bias. Based on a conservative strategy, this study proposes two new Q-matrix correction methods (SLR-B and eLasso). Under three general cognitive diagnosis models (GDM, LCDM, GDINA), the performance of the proposed methods is compared with existing methods. Simulation results show that the new methods, SLR-B and eLasso, exhibit superior performance in four metrics: attribute classification accuracy, pattern classification accuracy, correct attribute retention rate, and incorrect attribute correction rate. Overall, these methods outperform existing ones, with a relatively better performance in correct attribute retention. Improving item quality, increasing sample size, extending test length, or reducing the proportion of Q-matrix mis-specification can all enhance the effectiveness of various Q-matrix correction methods, with the impact of item quality being the most significant. Empirical analysis further supports the effectiveness of the new methods in Q-matrix correction, especially under the LCDM and GDINA models.