Research Forum: Using Item Response Theory to Refute Placement Decisions

This study explores the use of Item l{esponse Theory (IRT) or Rasch analysis in
making placement decisions. The general principles underlying populationdependent
classical theory standard analyses (including standard error of measure)
and population-independent lRT analyses are compared and are used to point
out the shortcomings of the classical analyses in making accurate placement
decisions. Two sets of hypothetical cut points based on raw scores and Raschgenerated
student ability estimates were applied to a set of data eN = 487) and
placement decisions using the two sets of cut points were compared. Twenty
discrepancies were found, meaning that five percent of the students were
potentially m~splaced when using their raw scores. This information may be
valuable for test administrators who want to make student placements based on
test results with the least amount of measurement error.