TABLE 6.

Differential item functioning test results for non-invariant items.

Item	Grouping variable	χ2	df	p FDR	UIDS	ESSD	Non-invariant parameters
4	Autism diagnosis	29.65	7	<0.001	0.290	−0.398	Slopes (a1 higher in AUT, a3 lower in AUT) Intercepts (all lower in AUT)
10	Autism diagnosis	19.90	6	0.003	0.211	0.260	Intercepts (d1 lower in AUT, d2–4 higher in AUT)
12	Autism diagnosis	37.63	6	<0.001	0.379	0.587	Intercepts (all higher in AUT)
16	Autism diagnosis	32.35	6	<0.001	0.358	0.374	Intercepts (all higher in AUT)
12	Sex	34.64	6	<0.001	0.420	0.650	Intercepts (all higher in Males)

Results indicate omnibus Wald tests of differential item functioning (DIF) using a version of the iterative anchor-selection method of Cao et al. (2017). Items presented in bold demonstrated differential item functioning large enough to be deemed “practically significant” (i.e., |ESSD| > 0.5). Parameter groups (i.e., either slopes or intercepts) that were significantly different between groups when tested alone with follow-up Wald tests (p < 0.05, uncorrected) are indicated in the “Non-invariant Parameters” column. Higher intercepts indicate less item difficulty (i.e., more item endorsement at a given latent trait level). UIDS, unsigned item difference in the sample (unsigned DIF effect size in response scale units); ESSD, expected score standardized difference (signed DIF effect size in Cohen’s d units); AUT, autism group.