Detail #

If we think an instrument measures two or more variables and that the variables are Nomothetic and consistently related to each other and to the items in the measure across people then CFA is a powerful way to see if, in a sample of scores on the instrument, the correlations between the items fit the expected pattern. For example for the Hospital Anxiety and Depression questionnaire CFA tests whether the seven anxiety items seem to relate to one shared source of difference (anxiety) across people and the seven depression items to a distinct source of difference (depression). This is a formal test of Construct Validity.

Where a measure only measures one thing, CFA can be applied to check unidimensionality: whether there does seem to be only one dimension of variation between individuals or if there appear to be more shared dimensions of variation than just the one. For therapy outcome/change measures true unidimensionality is highly unlikely so the question of real interest is how much variation appears to be affecting responses to all the items of the measure … which overlaps with Internal Reliability.

Try also … #

Factor Analysis (CFA), Exploratory Factor Analysis (EFA), Principal Component Analysis (PCA), Construct Validity, Multi-Dimensional Scaling (MDS).

Chapters #

Do we need this?

Further reading #

Nothing here yet!

Online applications #

Nothing here yet!