The Second Workshop on Statistical Methods in Cognitive Assessments

Thursday, May-16, Afternoon Session [2:00 5:35] Lihua Yao Multidimensional Item Response Theory: Applications and BMIRT, LinkMIRT, and SimuMIRT Software BMIRT (Yao, 2003) is a computer program that uses the Markov Chain Monte Carlo (MCMC) method to estimate item and ability parameters in the multidimensional multigroup IRT framework; exploratory and confirmatory approaches are supported. LinkMIRT (Yao, 2004) is a linking software that links two sets of item parameters onto the same scale in the MIRT frame work. SimuMIRT is software that simulates data for various MIRT models.Data requirements and formats, and sample data and input files will be provided to participants prior to the workshop; participants are recommended to go to www.BMIRT.com to download software into the laptop computers that they are required to bring to the workshop. Java run time environment (JRE), which can be downloaded on the internet, is required to be installed into the computer, and the computer path need to be set up following the directions in ”1.5.1 Java Run Time Environment” in the document BMIRT2013.pdf. Friday, May-17, Morning Session [9:10 9:55] William Stout A Family of Generalized Diagnostic Classification Models for Multiple-Choice Option-Based Scoring (GDCM-MC), with Applications to RUM & DINA Modeling A new family of restricted latent class diagnostic classification models (DCMs) for multiple choice item based assessments is presented, denoted by GDCM-MC. Its purpose is to allow researchers and practitioners to more fully capture the useful and fine-grained diagnostic information in a multiple choice item based diagnostic classification assessment, especially when there is useful diagnostic information in which incorrect options students choose. The primary targeted area of application is formative assessment, although summative assessment applications can also occur. In this regard, formative assessments are beginning to be designed where specific misconceptions (which students and teachers would like to correct) make choosing certain distractor options probable. We consider two important specific instances of the broad GDCM-MC family. These are an extended version of the Reparameterized Unified Model (sometimes called the Fusion Model), denoted ERUM-MC,