The Effect of Different Strategies for Combining Disordered Thresholds on Rasch Model Fit

Ömer Faruk DADAŞ; Derya Gökmen; Timur KÖSE

doi:10.5336/biostatic.2019-72009

The Effect of Different Strategies for Combining Disordered Thresholds on Rasch Model Fit

Articles > Article · 2020

Authors: Ömer Faruk DADAŞ, Derya Gökmen, Timur KÖSE

Turkiye Klinikleri Journal of Biostatistics, Other Refereed National Journals, National, TR DİZİN (ULAKBİM), 2020, vol 12, pp 53-69

DOI: 10.5336/biostatic.2019-72009

Summary

Objective: The internal construct validity of the scales is examined by Rasch analysis. There are limited number of studies on how to combine disordered thresholds in Rasch analysis. In this study, new four different category-combining strategies are proposed in order to combine disordered thresholds. The effects of these strategies on the overall fit of the Rasch model are investigated. Material and Methods: The strategies are obtained by combining the categories with disordered thresholds, 'to the left, to the right, to middle-left, and to middle-right'. It is decided to use the Partial Credit Model as the appropriate Rasch model. The data is obtained using the measurement tool, which assesses the disability levels of patients with rheumatoid arthritis. The internal construct validity of this scale is evaluated using RUMM2030 software. Results: In the measurement tools obtained after the category combining, information loss occurred due to the combining of the categories and it has been seen that standard errors of the ability of individuals have risen. However, the model fit and internal consistency of measurement tools have increased after the combining strategies. Conclusion: The category combining strategies are carried out on only one sample data, and since the single strategy is applied for all items in the combining process, the size of the disordered thresholds is ignored. Therefore, it is necessary to carry out a simulation study in order to generalize the results obtained.

View details