DINA 모형에서 응시생 분류방법의 분류 정확성 비교 연구 (A Note on Comparing Classification Accuracy of Examinee Classification Methods under DINA Model)

서지정보

· 발행기관 : 학습자중심교과교육학회
· 수록지 정보 : 학습자중심교과교육연구 / 13권 / 5호 / 121 ~ 142페이지
· 저자명 : 김지효

초록

본 연구의 목적은 DINA(deterministic-input, noisy “and” gate)모형에서 실제자료의 문항모수와 Q행렬을 이용하여 최대우도(maximum likelihood: ML), 최대사후확률(maximum a posteriori: MAP), 사후기대(expected a posteriori: EAP)방법들의 분류 정확성이 어느 정도인가를 알아보는 것이다. 연구를 위해 먼저 기초학력 향상도 평가 6학년 국어, 사회, 수학, 과학, 영어 과목에 대한 Q행렬을 개발하고 문항모수를 추정했다. 실제자료에서 얻은 문항모수와 Q행렬을 이용하여 참 인지요소(true )와 ML, MAP, EAP방법으로 추정된 인지요소간 정확일치도를 산출했다. 연구결과 모든 과목에서 ML, MAP방법보다 EAP방법의 정확일치도 평균이 높았다. 또한, 모든 인지요소를 동시에 정확하게 추정하는 방법은 MAP방법, 대부분의 인지요소를 정확하게 추정하는 방법은 EAP방법, 극심한 불일치를 피할 수 있는 방법은 EAP방법으로 나타났다.

영어초록

The cognitive diagnosis model(CDM) shows whether learners master specific attributes that items require. This specific information about the states of the attribute mastery can be used for teaching and learning because the theory identifies their strengths and weaknesses and diagnoses learner’s achievement levels precisely. In CDM, maximum likelihood (ML), maximum a posteriori (MAP), and expected a posteriori (EAP) methods are utilized to classify the examinees into the latent classes. It is important to examine which classification methods classify the examinees accurately. The important contribution of this study is to provide the guidelines for practitioners so that they can choose a better classification method.
In the study using item parameters and Q matrix from the real data, the Q-matrix was developed by teachers for Daejeon Tests of Basic Skills (DTBS) on 6th grade and the subjects were korean, math, social studies, science, and english. The percent of exact agreements between true skill pattern (true ) and estimated skill pattern by the three methods were calculated. The accuracy of the classification methods was assessed by the proportion of examinees classified correctly for all K attributes, the proportion of examinees classified correctly for K-1 attributes, the proportion of examinees classified incorrectly for K-1, or K.
The summary of the main results of this study is as follows:First, under the DINA model, when using item parameters and Q-matrix of DTBS, the EAP method showed relatively higher average in the percent of exact agreement than the ML and MAP methods across all subjects (korean, social study, math, science, and english) studied.
Second, All subjects MAP classified examinees correctly on all K attributes at a higher rate than ML, EAP. on the other hand, EAP classified more examinees correctly on at least K-1 attributes and produced fewer severe misclassifications.

참고자료

· 없음