NotesFAQContact Us
Search Tips
ERIC Number: ED318758
Record Type: Non-Journal
Publication Date: 1990-Apr
Pages: 47
Abstractor: N/A
Reference Count: N/A
Unidimensional Data from Multidimensional Tests and Multidimensional Data from Unidimensional Tests.
Reckase, Mark D.
Although the issue of dimensionality of the data obtained from educational and psychological tests has received considerable attention, the terms "unidimensional" and "multidimensional" have not been used very precisely. One use of the term dimensionality is to refer to the number of hypothesized psychological constructs believed to be required for successful performance on a test; this is "psychological dimensionality." A second use is to refer to the minimum number of mathematical variables needed to summarize a matrix of item response data; this is "statistical dimensionality" (SD). This paper presents a framework for discussing the SD of dichotomous data matrices of item response data. It is suggested that dimensionality refers to the number of dimensions needed to summarize a data matrix that results from interactions of a set of test items and a group of examinees. Two cases are presented to summarize the circumstances under which a set of test items will generate unidimensional data when responded to by a population of individuals. In the first case, all items are sensitive to the same combination of skills in the same manner. The second situation is when psychological dimensions are strongly confounded with the difficulty of the test items. Fourteen perspective, contour, or vector plots illustrate the discussion. (SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: Office of Naval Research, Arlington, VA.
Authoring Institution: N/A