I'm going to analysis my data in PLS-SEM and I want to know if it's necessary to do normality test. as we know, the result from PLS-SEM does not effect by the distribution of the data, therefore what is the benefit to do normality test ?!
Can you please comment me about this issue.
I appreciate your comments and thanks in advance
Dear Mohameed,
In SEM, a normal data has a potential if you undergo modeling since it will reduce your problem in parameter estimates (e.g. Standard errors). Though non-normal data doesn't affect much to your result with respect to present calculation methods such as ML. However, if you will know that your data is normal it will lessen your doubt about your data if your result is unsatisfactory with respect to your objective or hypothesis.
Dear Bahaa,
As you perhaps know that normality is not a requirement for using PLS-SEM. However, due to its "too much flexibility" regarding normality and other issues, PLS-SEM suffers from serious criticisms when it is compared with CB-SEM (please see the attached file). In fact, if you do normality test, there is nothing wrong and if your data "pass" the normality test, why not you should use CB-SEM to get more valid and accepted result?
Maybe no need to use normality for PLS-SEM but you should be able to acclaim that your findings are from normality data. I recommend you to test respective normality which is better than you do not.
Better to use Normality test along with multivariate assumptions. It will add value in your analysis and may yield more better and concrete results.
Dear Asmawil
thank you for your comment
actually I don't know my data normal or not, but i have justification for using PLS, so that why I'm not sure if have to normality test or not.
Dear Dr Taufique
thank you so much for your explanation and sending me the value article to spurt my argument.
my data is only 95 and my model a little complex , I also discussed with my supervisors about which software more appropriate with my study and finally we select PLS SEM depend on the study model, relationships, and hypothesis.
thankyou so much for your suggestion.
Non normal data have its own consequences , irrespective of the technique you are using. The data should not be too much away from normality in any case.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling