ERIC Number: ED206642
Record Type: Non-Journal
Publication Date: 1981-Apr
Reference Count: N/A
To Use or Not to Use--(The One- or Three-Parameter Logistic Model) That Is the Question.
Reckase, Mark D.
Definition of the issues to the use of latent trait models, specifically one- and three-parameter logistic models, in conjunction with multi-level achievement batteries, forms the basis of this paper. Research results related to these issues are also documented in an attempt to provide a rational basis for model selection. The application of the latent trait models is evaluated in terms of: (1) the assistance they can lend to test construction; (2) their use in the vertical equating of test scores; and (3) their use in comparing score scales. It is suggested that the target information function of the latent trait models can give control over test construction and precision, especially with the three-parameter model; but, there appears to be no acceptable way to select test items using the one-parameter model. The latter model is criticized for perpetuating the idea that all items are equal. Even when the test battery is sorted into unidimensional subtests, it is reported that neither the one- nor the three-parameter model is adequate for vertical equating purposes. It is concluded that both models have advantages and disadvantages. The recommendation that both models be used in conjunction with traditional procedures is suggested. (AEF)
Publication Type: Speeches/Meeting Papers; Information Analyses; Reports - Evaluative
Education Level: N/A
Sponsor: Office of Naval Research, Arlington, VA. Personnel and Training Research Programs Office.
Authoring Institution: Missouri Univ., Columbia.