Good fit to a one-dimensional IRT model is empirical evidence that the observed item responses can be explained by one continuous underlying trait. When it further can be concluded that the scale is meaningful (based on item analysis and association with external measures to assess its validity), and the assumption of measurement invariance across different subpopulations is tenable (Lubke et al. 2004), the approach effectively deals with non-normal distributions of sum scores in for instance psychopathology (Van den Oord et al. 2003).
