Measurement properties of the 12-item Short Form Health Survey version 2 in Australians with lung cancer: a Rasch analysis

Table 4 Overall Rasch model fit statistics and reliability of the SF-12v2^a

	Ideal	PCS-12	PCS-12 (subtest analyses)^d	MCS-12
Total item-trait interaction
Total item χ²		107.0	38.1	35.1
df		18	12	24
p value	> 0.05	< 0.001	< 0.001	0.067
Normed χ^{2 b}	≤ 2.5	5.94	3.18	1.5
Items
Fit residual (mean)	0	-0.50	0.05	0.10
Fit residual (SD)	< 1.5	3.36	2.07	1.15
Persons
Fit residual (mean)	0	-0.46	-0.38	-0.34
Fit residual (SD)	< 1.5	1.19	0.94	1.09
Uni-dimensionality
Equating t-tests Binomial dimensionality test (95% CI) Person-item residual correlation	< 0.05 (lower limit < 0.05) < 0.2	0.06 (0.03, 0.08) > 0.2 for items 2 and 3 > 0.2 for items 4 and 5	0.04 – < 0.2 for all items	0.02 – < 0.2 for all items
Person separation index^c	> 0.7	0.83	0.72	0.81
Equivalent Cronbach’s α	> 0.7	0.85	0.73	0.83

^aAs analysed using RUMM2030 (Rumm Laboratory Pty Ltd., Perth) for Windows
PCS-12, physical component summary; MCS-12, mental component summary; SD, standard deviation; CI, confidence interval; df, degrees of freedom
^bRatio of χ² value to degrees of freedom
^cRasch based reliability statistic (analogous to Cronbach’s α)
^dSubtest analyses for PCS-12 combining items 2 and 3 as well as items 4 and 5

ISSN: 1477-7525