NotesFAQContact Us
Search Tips
ERIC Number: ED453230
Record Type: Non-Journal
Publication Date: 2001-Apr
Pages: 31
Abstractor: N/A
Reference Count: N/A
Dimensionality, Dependence, or Both? An Application of the Item Bundle Model to Multidimensional Data.
Xie, Yiyu
A data set collected by the Science Education for Public Understanding Program (SEPUP) that has both dimensionality and dependence characteristics was studied, and an item bundle model nested in a multidimensional random coefficients multinomial logit (MRCML) model (R. Adams, W. Wang, and M. Wilson, 1997) was applied to this data. Data are from the field test of an assessment system for a yearlong middle school science curriculum, "Issues, Evidence, and You" (IEY). The SEPUP link tests have both multidimensionality and item dependence issues. Because the performance assessment was time consuming, only a small number of items could be given to each student during the testing period. For this reason, each item was specifically designed to be multidimensional and scored on a number of variables or elements. A multidimensional item bundle analysis suggests that most of the items are dependent within each bundle, no matter whether the bundles are unidimensional or multidimensional. Analyses showed that interaction effects did exist for most of the pairwise score combinations. The effects were more prominent for the bundles in which items measure content knowledge. Taking the item dependence into account may make the correlation among latent dimensions more accurate and meaningful. Nine appendixes contain additional information about the analyses. (Contains 18 references.) (SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A