Could you clarify your question a bit more? Are you developing a new scale? If so, and you want to evaluate your scale validity, you will need to collect additional variables for correlation analysis with your new scale.
Yes, Im developing a new scale. You mean I should use similar scales considering my conceptual framework in order to measure convergent and discriminant validity?
The answer to all methodological questions is "it depends"...
If you want to create and validate a shorter (or faster, cheaper, more acceptable) version of an existing validated scale, you could evaluate criterion validity by also administering the existing accepted long scale as your criterion. If the long and short scales measured a state over the same time interval it would be concurrent (criterion) validity. If you want your scale to have predictive properties you would need to collect data on the predicted outcome and you would be assessing predictive (criterion) validity.
If you are creating a scale in an area where there aren't any accepted validated scales, you would need to address construct validity, but to do this you don't administer other scales of the same construct (that is criterion validity, above). Evidence for convergent (construct) validity would be high correlations between your scale scores and a measure of something predicted by your theory to be associated with your construct (for example, physical activity -> BMI, or depression -> suicide attempts), and evidence for divergent (construct) validity would be low correlations between your scale and something not predicted to covary with your construct (e.g. physical activity -> left or right handedness, or depression -> eye colour).
Oh I got it now, I am now standing in new scale where there is no validated scales with that particular area i.e., cultural personality (religion factor). In this sense, I need to address only construct validity. But for test of convergent validity, as far as I know I need to exploit;
Standardized Factor Loadings (SFL) (>.7)
Average Variance Extracted (AVE) ( >.5)
Construct Reliability (CR) (>.7)
for test of divergent validity, what should I measure?
Thanks ahead
Now, after some times, I am aware that it is must to use other related scales to validate convergent and discriminant that is essential in validation study. I guess you are all agreed with me.
- Badges
- Science topic
- Similar topics
- Publishing