Related Experiment Videos
A critique of Rasch residual fit statistics
1Department of Biometry and Genetics, Louisiana State University Medical Center, 1901 Perdido Street, Box P5-2 New Orleans, LA 70112-1393, USA. gkarab@lsumc.edu
Journal of Applied Measurement
|May 25, 2002
Summary
Rasch model fit statistics are crucial for objective measurement but residual fit statistics have faulty properties. Improved analysis requires residual-free statistics or optimal indices for detecting measurement disturbances.
Area of Science:
- Psychometrics
- Educational Measurement
- Statistical Modeling
Background:
- The Rasch model is vital for objective measurement of abilities and item difficulties.
- Data fit to the Rasch model is essential for achieving measurement objectivity.
- Fit statistics are critical for identifying data inconsistencies that threaten objectivity.
Purpose of the Study:
- To critically analyze the measurement quality of residual fit statistics in Rasch model analysis.
- To highlight the complexities and limitations of traditional Rasch fit analysis.
- To propose improvements for more accurate detection of measurement disturbances.
Main Methods:
- Critical analysis of residual fit statistics for the Rasch model.
- Examination of the statistical properties of residual fit statistics.
- Discussion of the impact of sample and test properties on fit statistic performance.
Main Results:
- Residual fit statistics possess faulty properties that complicate Rasch fit analysis.
- A single critical value for misfit diagnosis is inadequate across diverse testing situations.
- Current methods lead to both overdetection and underdetection of misfit.
Conclusions:
- Rasch fit analysis is more complex than commonly perceived.
- Psychometricians should adopt residual-free fit statistics or statistically optimal indices.
- Improved fit statistics are necessary for reliable measurement objectivity in Rasch modeling.