Figure 1.

Multi-paneled figure, with different kinds of information about the items on the y-axis, and the item numbers on the x-axis. The x-axis is ordered by the results of the Rasch mixture model analysis, with the smallest DIF on the left (item 2), and the largest DIF on the right (item 30). (a) Severity of DIF, where a higher DIF indicates larger differences between groups, as measured in three analyses: Rasch mixture models (triangles), Rasch regression trees (circles), and traditional methods (crosses). (b) Item characteristics, such as, whether the item was negatively phrased (Negative, in empty squares), contained a word denoting frequency (Frequency, in filled squares), contained the word “very” (in empty circles), contained a word denoting difficulty (Difficulty, in filled circles), contained a comparison (Compare, in empty triangles), and/or was a reversely coded item (Recoded, in filled triangles). (c) Whether the items were included in AQ-28 and AQ-10. (d) Whether the items were identified as misfitting, or as showing DIF, in a previous study comparing groups. The items with faint coloring were included in the versions used in these studies. Lundqvist and Lindner (2017) used all 50 items.

Grove: Grove et al., (2017); Murray’14: Murray et al. (2014); Murray’17a: Murray, Allison, et al. (2017); Murray’17b: Murray, Booth, et al. (2017); L&L: Lundqvist and Lindner (2017); MF: misfit; DIF: differential item functioning; AQ: Autism Spectrum Quotient.