Detecting differential item functioning using generalized logistic regression in the context of large-scale assessments

Large-scale Assessments in Education

An IERI – International Educational Research Institute Journal

Table 4 Average power rates across studied conditions for 10 and 20 groups

*Panel (a) Invariance in Discrimination Item Parameters*
		Difference in Parameters for DIF Items
	% of DIF items	Δ_a = .40% of Groups with DIF Items				Δ_a = .80% of Groups with DIF Items
		40	70	40	70	40	70	40	70
		10 Groups		20 Groups		10 Groups		20 Groups
Discrimination	20	.70	.51	.70	.55	1.00⁺	.99	1.00	1.00
	20	(.08)	(.07)	(.07)	(.07)	(.01)	(.02)	(.00)	(.01)
	40	.65	.57	.64	.61	.99	.98	.99	.98
	40	(.05)	(.04)	(.03)	(.04)	(.01)	(.01)	(.01)	(.01)
	60	.71	.63	.72	.68	.97	.97	.97	.97
	60	(.04)	(.04)	(.03)	(.03)	(−)	(.01)	(.01)	(.01)
*Panel (b) Invariance in Difficulty Item Parameters*
		Difference in Parameters for DIF Items
	% of DIF items	Δ_b = .50% of Groups with DIF Items				Δ_b = 1.00% of Groups with DIF Items
		40	70	40	70	40	70	40	70
		10 Groups		20 Groups		10 Groups		20 Groups
Difficulty	20	.97	.92	.95	.95	1.00	1.00	1.00	1.00
	20	(.03)	(.03)	(.01)	(.03)	(.00)	(.00)	(.00)	(.00)
	40	.98	.96	.98	.96	.98	.98	.98	.98
	40	(.02)	(.02)	(.02)	(.02)	(.01)	(.01)	(.01)	(.01)
	60	.20	.20	.18	.18	.20	.19	.18	.18
	60	(.02)	(.02)	(.02)	(.02)	(.02)	(.02)	(.01)	(.01)

Note. Values in () represent SDs per condition; (−) SD was rounded to zero (actual value .004); + = rounded to 1.00.