Hi everyone, i´m looking for documentation about convergent and discriminant validity. Specially i want to understand the right way to judge if two measures are correlated. Thanks...
Juan Pablo Sánchez, I'm sorry that nobody has stepped in to offer some suggestions for you - so I hope it's not too late for me to do so now.
I have found it impossible to find anything that I regard as definitive or satisfactory in the literature, but my personal view is that the absolute value of correlations for convergent validity should ideally range from .40 to .60 (i.e., +.40 to +.60 OR -.40 to -.60).
If the correlations are more extreme that that (i.e., if they move above .60 toward 1.00, or below -.60 toward -1.00), they can indicate an undesirable degree of conceptual overlap between the two constructs of interest, suggesting that the same construct (or its inverse) is simply being measured by two different procedures rather than that different, but related, constructs are being examined.
As far as discriminant validity is concerned, I think that correlations between -.15 and +.15 are desirable. Maybe in the range -.20 to +.20 would be OK.
If you or anyone else found citations that are relevant to the above, I'd appreciate knowing about them.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size