Levine and Rubin (1979) used the term appropriateness measurement to denote methods for detecting persons with a so-called aberrant item response pattern on a test or questionnaire. Aberrant is used here in the sense that the item response patterns are not corresponding to those one would typically expect to observe. Appropriateness measurement seeks to provide internal evidence indicating that the person did not approach the test as do other test-takers, and this solely based on the item responses given, not referencing other additional information not part of the test itself (e.g., a person’s gender, religion, race, or socio-economic status).
Assessing measurement appropriateness is relevant to reach more valid test results and inferences in general, but especially to prevent making important decisions about individuals when measurement of that individual is in fact shown to be inappropriate.
In this seminar, Braeken will:
- give a brief introduction to this subfield of psychometrics that connects to general methodological areas such as assessment of local model fit (i.e., person fit) and multivariate outlier detection.
- and discuss recent work on mixture model expansions to reduce calibration bias in the assessment of measurement appropriateness.