Table 3.

Summary fit statistics from Rasch analysis

Version	Data changes	Sample size	Item-trait interaction χ2		Item fit residual		Person fit residual		PSI	Significant t-tests
			Value (df)	P value	Mean	SD	Mean	SD
Initial	None	145	73.512 (13)*	0.000	0.49	1.84	−0.22	1.18	0.89	7.0
Run 2	Deleted 8 misfit persons	137	72.550 (13)	0.000	0.43	1.93	−0.14	1.01	0.90	5.2
Run 3	Rescored all misfit items	137	19.693 (13)	0.103	−1.01	1.21	−0.66	1.29	0.89	1.6
Run 4a	Deleted 20 misfit persons	117	16.105 (13)	0.243	−0.58	1.12	−0.41	1.08	0.90	4.8
Run 5	Formed 4 testlets	117	0.937 (4)	0.92	0.26	0.72	−0.35	0.87	0.84	1.3
Run 6a	Used initial data, rescored all items	145	19.480 (13)	0.108	−1.07	1.33	−0.70	1.33	0.89	4.3
Run 7	Formed the 4 testlets again	145	3.546 (4)	0.47	0.02	0.85	−0.45	0.89	0.83	2.9
Run 8	Deleted 1 misfit persons	144	3.612 (4)	0.46	0.03	0.85	−0.43	0.86	0.82	2.9
Rasch-Refined	Deleted 3 persons with incomplete data	141	3.613 (4)	0.46	0.00	0.87	−0.44	0.86	0.82	3.0

Criteria of fit to Rasch Model: minimum sample size of n = 108, PSI (Person Separation Index) ≥ 0.80 for reliability assessment by measurement scale, χ² P-value > 0.05 [Bonferroni-adjusted], Items- and Persons- Fit Residual Mean ~ 0 and SD (Standard Deviation) ~ 1, less than 5 % significant t-test

*Significant after P < 0.05 with Bonferroni correction for the number of items in the analysis

^aHad local item dependency