A new scale has recently been developed to assess quality of life. The WHOQL is another well-established scale. How can we compare these two in terms of psychometric properties to see if the new scale is better? Notably, a four-factor solution is suggested for both, with conceptually similar factors (psychological, physical, social, and environmental).
So far, I have thought of several methods:
- to compare the fit of the structure using the fit indices such as CFI, SRMR, RMSEA (Not sure if this is correct, but we may use bootstrap to compare these indices across questionnaires)
- To compare the structure in terms of cross-loadings. That is, we can fit a traditional CFA model and a model that allows for cross-loading (an Exploratory Structural Equation Model [ESEM]). Then the two questionnaires can be compared in terms of reduction in fit when the cross-loadings are fixed to zero. We can also calculate the mean (or median) item complexity for both and compare them (even statistically using bootstrap). Also, we can calculate a correlation matrix showing how the intercorrelation among latent factors is affected by fixing the cross-loadings to zero (i.e., the latent correlation matrix from ESEM minus that of CFA).
- To compare the reliability indices (even statistically using bootstrap).
- To compare them in terms of length or readability
- To compare them in terms of missingness
- To compare the item contents
Any feedback on these thoughts or any related reference or any other suggestion will be appreciated.
Thank
Just to add to your suggestions, you can also check the reliability of both the scales (Alpha value).
You might want to consider methods using the various sources of validity:
- Relations to other variables: Examine the correlations of both measures to other variables that are theoretically related and unrelated. Using the convergent/discriminant continuum, this can help you understand how the two measures to other measures are similar and different from what you would expect.
- Response processes: How do individuals interpret the items of each measure and do these match what the test developer intended?
- For item content, have subject-matter experts and target population both rate the items.
s. Rama Gokula Krishnan , you mentioned alpha. Given the question discusses four factors, would one of the omega statistics be better suited?
Please, send me an example for compearing two questionnairies by CFA. The questionnairies are with different length and in the first instrument I have 7 scales - in the second - 3 scales. The both measure the emotional sensitivity of the pupils in secondary school.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology