ABSTRACT
Chapter 12 of ARM 4th Ed is entitled Rasch Model Requirements: Fit & Unidimensionality. The most distinctive and contentious aspects of Rasch measurement are that the collected data should “fit the model” and that unidimensionality is a prerequisite of scientific measurement. The chapter details the concepts behind residual-based fit statistics and the PCA of Rasch residuals. Four key Rasch fit statistics are introduced and issues related to overfit and underfit of data to the Rasch probabilistic model are canvassed. Methods and techniques for detecting and handling multidimensionality are demonstrated and discussed. Interpretation of fit statistics for aberrant item and person performances are introduced and the problem of guessing is further examined.
