PARTIALLY OBSERVED MIXTURES OF IRT MODELS: AN EXTENSION OF THE GENERALIZED PARTIAL CREDIT MODEL

The generalized partial credit model (GPCM) is used frequently in educational testing and in large-scale assessments for analyzing polytomous data. Special cases of the generalized partial credit model are the partial credit model—or Rasch model for ordinal data—on the one hand and the 2-parameter logistic (2PL) model on the other hand. The extension proposed here enables the use of the GPCM in discrete mixture item response theory (IRT) models with partially observed mixture information. Partially observed mixtures are discrete mixture distributions with missing grouping information for some, but not all, observations. This concept of partially observed mixtures integrates multigroup and mixture distribution IRT models. The proposed model includes the latent class analysis (LCA) and (multigroup) IRT models as well as mixture distribution IRT models as submodels. An application of the proposed partially observed mixture IRT model is presented. This application classifies about 80% of data from a large-scale mathematics assessment into three school types and estimates a mixture IRT model for the whole sample in which the grouping information is unknown for 20% of the sample. The aim of the method is twofold: first, to estimate the latent distributions by school type and, second, to unmix the data with missing school types into the groups established by the portion of the sample with known school-type information.

[1]  M. R. Novick,et al.  Statistical Theories of Mental Test Scores. , 1971 .

[2]  G. Masters A rasch model for partial credit scoring , 1982 .

[3]  A. Formann Linear Logistic Latent Class Analysis for Polytomous Data , 1992 .

[4]  G. Schwarz Estimating the Dimension of a Model , 1978 .

[5]  Richard Goldstein Latent Class and Discrete Latent Trait Models: Similarities and Differences , 1998 .

[6]  R. D. Bock,et al.  Marginal maximum likelihood estimation of item parameters , 1982 .

[7]  Jürgen Rost,et al.  A logistic mixture distribution model for polychotomous item responses , 1991 .

[8]  H. Akaike,et al.  Information Theory and an Extension of the Maximum Likelihood Principle , 1973 .

[9]  E. Muraki A Generalized Partial Credit Model: Application of an EM Algorithm , 1992 .

[10]  J. Rost,et al.  Applications of Latent Trait and Latent Class Models in the Social Sciences , 1998 .

[11]  John Uebersax,et al.  Statistical Modeling of Expert Ratings on Medical Treatment Appropriateness , 1993 .

[12]  Robert J. Mislevy,et al.  Modeling item responses when different subjects employ different solution strategies , 1990 .

[13]  Kentaro Yamamoto,et al.  Chapter 4: Item Response Theory Scale Linking in NAEP , 1992 .

[14]  Matthias von Davier,et al.  Polytomous Mixed Rasch Models , 1995 .

[15]  Georg Rasch,et al.  Probabilistic Models for Some Intelligence and Attainment Tests , 1981, The SAGE Encyclopedia of Research Design.

[16]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[17]  T. Heinen,et al.  Latent Class and Discrete Latent Trait Models: Similarities and Differences , 1996 .

[18]  R. D. Bock,et al.  Marginal maximum likelihood estimation of item parameters: Application of an EM algorithm , 1981 .

[19]  E. Muraki A Generalized Partial Credit Model , 1997 .

[20]  Kentaro Yamamoto,et al.  Item Response Theory Scale Linking in NAEP , 1992 .

[21]  Robert J. Mislevy,et al.  Estimating Population Characteristics From Sparse Matrix Samples of Item Responses , 1992 .

[22]  Raymond J. Adams,et al.  Multilevel Item Response Models: An Approach to Errors in Variables Regression , 1997 .

[23]  Kentaro Yamamoto,et al.  HYBRID MODEL OF IRT AND LATENT CLASS MODELS , 1982 .

[24]  Jürgen Rost,et al.  Rasch Models in Latent Classes: An Integration of Two Approaches to Item Analysis , 1990 .

[25]  Bradley A. Hanson,et al.  Estimation of Item Response Models Using the EM Algorithm for Finite Mixtures. , 1996 .