The construction of two-stage tests

Jos J. Adema

Research output: Book/ReportReportOther research output

11 Downloads (Pure)

Abstract

Although two-stage testing is not the most efficient form of adaptive testing, it has some advantages. In this paper, linear programming models are given for the construction of two-stage tests. In these models, practical constraints with respect to, among other things, test composition, administration time, and inter-item dependencies play an important role. Research indicates that two-stage tests can be constructed both sequentially and simultaneously. Models based on the maximin model for test construction developed by W. J. van der Linden and E. Boekkooi-Timminga (1989) are formulated for the sequential case, with constraints specified at test and subtest levels. It is assumed that a bank of items calibrated under the item response model is available and that "information" is used in accordance with G. H. Fischer's information model. The maximin design is used in order to select the items that maximize the information in the test, while the resulting test information function still has the desired shape. The paper concludes that simultaneous test construction has the disadvantage of having a large number of variables; hence, constraints must be considered.
Original languageEnglish
Place of PublicationEnschede, the Netherlands
PublisherUniversity of Twente, Faculty Educational Science and Technology
Number of pages34
Publication statusPublished - 1988

Publication series

NameOMD research report
PublisherUniversity of Twente, Faculty of Educational Science and Technology
No.88-14

    Fingerprint

Keywords

  • Latent trait theory
  • Test construction
  • Linear programming
  • Item banks
  • Mathematical models
  • Mathematical formulas
  • Adaptive testing
  • Computer Assisted Testing

Cite this

Adema, J. J. (1988). The construction of two-stage tests. (OMD research report; No. 88-14). Enschede, the Netherlands: University of Twente, Faculty Educational Science and Technology.