I have a questionnaire with questions that measure one variable. The questions have the Likert scale format. I want to check convergent or discriminant validity for the questionnaire. How can I carry out it?
You need to have measures from other surveys/questionnaires and/or behavioral tasks. For instance, if you have two other measures, one of which is supposed to correlate with your current scale score (in theory, at least), you ought to get a significant correlation (or significant beta coefficient, if you execute regression analysis). And if it does not correlate with the second measure that is NOT supposed to correlate with it, you will get evidence for convergent and discriminant validity. So in a nutshell, convergent and discriminant validity go hand in hand, you need to show that your pre-existing scale score correlate with another measure A (evaluating the same underlying construct) but not measure B (evaluating a different construct).
Hope this helps!
As Jimmy Y. Zhong says, you need to have additional measures where you can predict degree and direction of the correlation with your new measure. For convergent validation, there should be a high positive correlation. For discriminant validity there should be a zero or negative correlation.
Concerning convergent validity you have to find the average variance extracted (AVE) of the latent construct. According to Fornell and Larcker (1981), "AVE measures the level of variance captured by a construct versus the level due to measurement error." Koufteros (1999) defend that any level above 0.5 is considered acceptable.
Concerning discriminant validity Fornell and Larcker (1981) define that it can be obtained when the amount of the variance captured by the construct is compared with the shared variance with other constructs.
You can use any statistical software like SPSS to measure both of them.
Good luck in your research!
can we find discriminant validity through method of contrasted groups..e.g., i m developing a diagnostic scale and take 2groups (1with problem and other without problem).and check the discriminatory power of scale in differentiating these two groups with problem and without problem....should i call it a discriminant validity??????
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire