Does anyone have a better ideas how to explain the used of Heterotrait-Monotrait Ratio of Correlations (HTMT) in assessing the discriminant validity in PLS-SEM model?
Dear Muhammad, I am sorry, I need clarification: better that what? In general, HTMT is a common approach to gain insights into discriminant validity and very useful - but only I you chose to compare traits that are not to far away from one another, otherwise the results become trivial. A more sophisticated approach would be the analysis of nomological networks not just by means of manifest correlations (like Pearson's r) but by latent structural equation models. I prefer the latter, however, also manifest structural equation models work well - they just tend to underestimate the relationship between two traits which might sometimes turn out to be a problem especially for discriminant validity.
Use the HTMT criterion to assess discriminant validity! If the HTMT value is below 0.90, discriminant validity has been established between two reflective constructs. If you like to obtain the HTMT_Inference results, you need to run the bootstrapping routine. When starting the bootstrapping routine, it is important that you select the option “Complete Bootstrapping”. Then, in the bootstrapping results report, you find the bootstrapped HTMT criterion results in the section “Quality Criteria”.
Take a look at the following video to see how to use SmartPLS to get HTMT results:
For detailed explanations of the HTMT criterion for discriminant validity assessment in variance-based structural equations modeling, see Henseler, J., Ringle, C. M., and Sarstedt, M. 2015. A New Criterion for Assessing Discriminant Validity in Variance-based Structural Equation Modeling. Journal of the Academy of Marketing Science, 43(1): 115-135. http://link.springer.com/article/10.1007%2Fs11747-014-0403-8
@arumugam, I have question. If the HTMT ratio calculated after the bootstrapping and it is found that the HTMT ratio between construct and sub-construct is more than 1, is it ok? Or HTMT ratio is only applied among constructs? Thank you
HTMT criterion measures the average correlations of the indicators across constructs. The acceptable levels of discriminant validity (< 0.90) as suggested by Henseler et al. (2015).
It is recommended for identifying distinctiveness of the constructs but should be used taking 0.85 or 0.90 cautiously. its limitations should be cared.
HTMT ratio must be less than 1.00 (Henseler, Ringle, & Sarstedt, 2015).
This article explains it well: Gaskin, J., Godfrey, S., & Vance, A. (2018). Successful system use: It’s not just who you are, but what you do. AIS Transactions on Human-Computer Interaction, 10(2), 57-81.
I am of the opinion that the acceptance and rejection criterion should be set because going by the position of Henry and Roshan a decision region is very important like John said when it's less than 1.00 but what of if it's greater than 1.00 as it was asked by Joachim. Please guide us on the decision rule of HTMT so that we the young researcher can learn from your conclusion. Thanks
HTMT > 0.9 or even >1 says that the latent constructs are not distinguishable, right? Does this mean; I have to resort the items along these two constructs like trial and error to make them more distinguishable? and what should I do, if this issue cannot be fixed by resorting the items? run a new EFA?
Uuuh, I guess I know why this happened. I developed a higher order construct out of my 4 latent constructs, where you have all the indicators of the latent constructs also in the higher order construct, when you do this in SmartPLS3. However I have some issues with HTMT > 0.9:-( need to resolve this
Sara Hoseingholizade , I found out why HTMT is > 1, because I used a higher order latent construct on the basis of a repeated indicator approach, see https://www.youtube.com/watch?v=LRND-H-hQQw&t=648s