Factors Affecting the Item Parameter Estimation and Classification Accuracy of the Cognitive Diagnostic Models


SÜNBÜL S. Ö. , KAN A.

HACETTEPE UNIVERSITESI EGITIM FAKULTESI DERGISI-HACETTEPE UNIVERSITY JOURNAL OF EDUCATION, vol.31, no.4, pp.778-795, 2016 (Journal Indexed in ESCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 31 Issue: 4
  • Publication Date: 2016
  • Doi Number: 10.16986/huje.2015014663
  • Title of Journal : HACETTEPE UNIVERSITESI EGITIM FAKULTESI DERGISI-HACETTEPE UNIVERSITY JOURNAL OF EDUCATION
  • Page Numbers: pp.778-795

Abstract

The purpose of this study is to investigate factors affecting the item parameter estimation, item fit and classification accuracy of the Cognitive Diagnostic Models (CDM). For this purpose, the data is generated by using noncompensatory model (DINA) according to various factors (sample size, correlation between attributes, the number of attributes, the number of item, s and g parameters levels). The simulated data were analyzed by using DINA models. Data simulation and analyses were conducted by using R 3.0 with CDM package. The output files were organized for parameter estimation, item fit and classification accuracy for both main and interaction effects. By using DINA analysis model obtained from mean values of "Absolute Mean Bias" (MOY) to estimate g parameter, sample size, number of items and levels of g and s parameters of significant effect were observed. s parameter estimation obtained from the mean values MOY, sample size, level of the correlation between attributes and level of the s and g parameters of a significant effect were observed. By using DINA analysis model obtained from mean values of RMSEA, sample size, number of item, number of attribute and levels of g and s parameters of significant effect was observed. By using DINA analysis model obtained from mean values of "Correct Classification Rate (CCR)", number of item, number of attribute and levels of g and s parameters of significant effect was observed.