University of Illinois Urbana-Champaign

New techniques for the dimensionality assessment of standardized test data

Kim, Hae-Rim

This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.

Permalink

https://hdl.handle.net/2142/19110

Description

Title
New techniques for the dimensionality assessment of standardized test data
Author(s)
Kim, Hae-Rim
Issue Date
1994
Doctoral Committee Chair(s)
Stout, William F.
Department of Study
Statistics
Discipline
Statistics
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Keyword(s)
  • Education, Tests and Measurements
  • Statistics
  • Psychology, Psychometrics
Language
eng
Abstract
  • A new index based on the conditional covariance of item scores given a latent variable is defined and investigated to assess dimensionality in educational and psychological test data. This index successfully detects the test dimensionality in both identifying the number of dimensions present in the test and identifying the items contributing to each dimension by means of cluster analysis. In addition, this index remarkably quantifies lack of unidimensionality in the test data, and the asymptotic behavior of this index under unidimensionality provides theoretical justification.
  • A new significance test based on a kernel smoothing technique, is developed to detect dimensional discrepancy of item pairs. In comparison to existing procedures a simulation study of this method reveals a reasonable type 1 error rate with respect to its nominal level of significance as well as excellent power performance.
  • The unidimensional parametric item and ability calibration procedures, BILOG and LOGIST, are examined to reveal what is actually being estimated as unidimensional ability when data is not unidimensional. Ability estimation accuracy is also investigated in terms of average standard error. Both BILOG and LOGIST appear to provide a composite of underlying latent traits as their presumed unidimensional ability estimate. In doing so, the average standard errors are relatively invariant to the degree of lack of unidimensionality, however the direction of the composite being measured best changes systematically and by a large amount, according with varying amount of multidimensionality.
Type of Resource
text
Permalink
http://hdl.handle.net/2142/19110
Copyright and License Information
Copyright 1994 Kim, Hae-Rim

Owning Collections

Graduate Dissertations and Theses at Illinois PRIMARY
Graduate Theses and Dissertations at Illinois
Dissertations and Theses - Statistics