Stepwise Item Selection Procedures for Rasch Scales Using Quasi - Loglinear Models

Two iterative procedures for constructing Rasch scales are presented. A log-likelihood ratio test based on a quasi-loglinear formulation of the Rasch model is given by which one item at a time can be deleted from or added to an initial item set. In the so-called "top-down" algorithm, items are stepwise deleted from a relatively large initial item set, whereas in the "bottom-up" algorithm items are stepwise added to a relatively small initial item set. Both algorithms are evaluated through a simulation study with generated data. Item parameters are given for four generated unidimensional data sets and two generated two-dimensional sets. Abilities were randomly sampled from a multivariate normal distribution with a sample size of 1,000. Results for the top-down algorithm were poor, but results for the bottom-up algorithm were more encouraging. It is suggested that alternating the bottom-up algorithm with one or two iterations of the top-down algorithm would allow the procedure to reject items that were added incorrectly in a previous step. Eight tables illustrate the item parameters and the use of both algorithms for the generated data. (SLD) *********************************************************************V* Reproductions supplied by EDRS are the best that can be made from the original document. **********A************************************************************ Stepwise Item Selection Procedures for Rasch Scales Using Quasi-Loglinear Models U S DEPARTMENT OF EDUCATION Once of Educabonat Research and imprmement EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC) TAtus document has been reproduced as 'mewed from the person or organaatton ongtnatmg a C Mmor changes have been made to improve reproduction quality Points of ,new or opimOnS stated m thisdoc umem do not necessanly represent othval OERI positron or policy