Table 2 Statistical tests and criteria for assessment of measurement properties of the SF-12v2

Dimensionality

Local dependency

Response to an item should not be dependent on the response to another item

Uni-dimensionality

Extent to which items for the PCS-12 and MCS-12 measure one underlying construct

Person-item residual correlation values > 0.2 above the average correlation is indicative of local dependency [22, 23, 30]

Uni-dimensionality confirmed if < 5% of significant t-tests between two most dissimilar subsets of items identified from the PCA of standardised residuals [22]. Where > 5% significant t-tests, uni-dimensionality supported if lower bounds of CI < 0.05 [22]

Response thresholds

Degree to which participants were able to discriminate between the response options for each item in the SF-12v2

Examination of pattern of thresholds from the threshold map

Inspection of category probability curves. Thresholds considered to be ordered if each response option systematically has a point along the location continuum to be the most likely response

Internal consistency

The degree of inter-relatedness among SF-12v2 items

Person separation index (PSI) values > 0.70 indicates good internal consistency (similar to Cronbach α values) [22]

Measurement invariance (item bias)

Whether or not different groups with similar characteristics (e.g. men vs women, public vs private) respond differently to a given item

Measured using differential item functioning (DIF). Uniform DIF is indicated by a significant main effect for the person factor (e.g. sex) using a Bonferroni adjusted p value for significance [22]. Non-uniform DIF is indicated by a significant interaction effect [22]

Targeting

Degree to which the PCS-12 and MCS-12 scores was targeted to patients with lung cancer

A well-targeted scale will have mean location logit score of zero [22]. Items will also be well-aligned on the person-item threshold distribution map