SmartPLS: What should I do if discriminant validity HTMT>0.90? Is any alternative available for SMART PLS?
Hello Garima,
Note that HTMT2 is mentioned as a possible improvement over HTMT as an indicator of whether latent variables are indeed distinct (see this link: https://henseler.com/htmt.html).
Regardless, if you determine that your latent variables are not sufficiently distinguishable (but you have strong reason to believe that they should be), your options are:
1. Revise or replace the indicator/manifest variables used for one or more of the factors;
2. If there is doubt that your sample is genuinely representative of the target population (or that the sample was too small), collect more/new data;
3. Rethink the basis for asserting that the factors should be distinguishable; perhaps a second-order or bi-factor model would more accurately correspond to what you really have.
As for alternatives to PLS-SEM, do you mean: (a) other PLS software; or (b) other methods for mustering evidence for convergent and discriminant validity?
Good luck with your work.
I will suggest you decrease the value of items it will automatically be corrected. For example, if one variable has five responses as [5,5,5,5,5] and another also has the same [5,5,5,5,5] you have to make a slight change in the responses by deleting respondent's reported similar responses.
Hello Garima,
You can use Fornell-Larcker criterion where if the square root of each construct's AVE is higher than its correlation with another construct, discriminant validity may be established.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Probability
- Probability Theory
- Reliability