Messy Middle or Messy Model: Challenges in Learning Progression Assessments

• In Event: Roundtable Session 46
  In Roundtable Session: Assessment and Learning Progressions: Designs and Other Considerations

Abstract

This research explores the set of conditions in which the ordering of the items by their b-parameters is altered when incorrectly, but undetectably, fıtting the Rasch model to the data. Specifically, this study aims to find the factors affecting this ordering when fitting a Rasch model to data generated with a 3PL model. We will show that using 3PL model rather than the Rasch or 1PL may result in better ability estimation even when the sample size is small, in which estimation of the guessing parameter may be challenging. The results of this study is expected to be used to overcome some of the learning progression related assessment problems and challenges.

Authors

• Lokman Akbay, Rutgers University

• Nathan D Minchen, Rutgers University - New Brunswick/Piscataway

• Jimmy de la Torre, Rutgers University