I knew that HTMT2 is some kind of improvised version of HTMT (Heterotrait-Monotrait ratio), and usually bootstrapping can be used to provide the 95%CI for the HTMT score in SmartPLS. As for HTMT2 (SmartPLS does not have this feature yet), I calculate it using the R-Studio or online calculator and neither of them can give me the confidence interval, how can I obtain the 95%CI for HTMT2 then ?
Reference: https://www.emerald.com/insight/content/doi/10.1108/IMDS-02-2021-0082/full/html
It's okay to use HTMT :) In fact, SmartPLS uses from its very beginning a third version (you may call it HTMT+), by computing the average absolute correlations: https://www.smartpls.com/documentation/algorithms-and-techniques/discriminant-validity-assessment
Since the original HTMT has issues with negative correlations and the so-called HTMT 2 only marginally differs in its results from HTMT+, there are unfortunately no plans to implement another HTMT version in SmartPLS.
Dear Anak,
You can find an implementation of the HTMT2 including the corresponding CI in the R package cSEM. The HTMT+ and the HTMT2+ are implemented there as well ;)
Best regards,
Florian
Thank you mr. Christian M. Ringle for your prompt answer. It is an honor. If I may ask you directly, is it true that as long as the upper bound of the confidence interval for this HTMT does not reach 1.0, then you can get a discriminant validity for the model? Regards.
Anak Agung Bagus Wirayuda , thank you very much. Here is a related discussion:
https://www.researchgate.net/post/Heterotrait-Monotrait-Ratio-of-Correlations-HTMT-in-assessing-the-discriminant-validity-in-PLS-SEM-model
This could be fine in accordance withArticle A New Criterion for Assessing Discriminant Validity in Varia...
But you may want to consider more conservative criteria based on the findings in this article: Article Heuristics versus statistics in discriminant validity testin...
Dr Christian M. Ringle , so even if my HTMT result is slightly above 0.85 or above 0.90, but if 95%CI is below 1.00, can I assume discriminant validity? Sorry for asking perpetually, I hope you do not mind. Thank you in advance.
Anak Agung Bagus Wirayuda : While the original paper says significantly below 1,
we suggest being stricter in the 3e of the PLS-SEM primer (i.e., significantly < 0.85 or 0.90) to be on the safe side and to ensure the validity of the results
--> Book A Primer on Partial Least Squares Structural Equation Modeli...
But make sure to use a one-sided test with a 5% or 10% probability of error.
- Badges
- Science topic