IMPROVEMENT ON DIAGNOSING EXAMINEES' ATTRIBUTES-MASTERY PATTERN USING BAYESIAN INFERENCE
LU Jing-ying · Journal of Beijing Normal University · 2012
This article improves Bayesian inference for binomial proportion(BIBP) which was proposed by Kim,H.S.J.(2011) to diagnose examinees' attribute mastery in rule space model The method updates the posteriors of single-attribute parameters by single attributes-mastery probability and response data to multiple-attribute items.Simulation method was used to generate item response data.Pattern match ratio and probability match ratio were used as criteria to evaluate classification accuracy of different approaches.Parameter estimation was obtained in three ways: with and without updating posteriors of single-attribute parameters by attributes-mastery pattern and attributes-mastery probability.Data showed that improvement of BIBP perform better classification effect than other methods.