To Ned Kock - heterotrait-monotrait ratio of correlations - HTMT - updates in WarPLS?
BRox
I hope you are keeping well. Thanks for the amazing progress with WARPPLS!
Since the seminal papers on the use of HTMT ratio of correlations as measure of discriminant validity came out a few years ago,editors and reviewers have been asking for it in journal articles.
Do you think WarpPLS can generate this HTMT statistic in the near future? In the meantime, what do you suggest to us WARPPLS users to use - an excel template maybe?
Thanks.
BROX
Ned Kock
Hi BROX.
The problem with any new method that is aimed specifically at a particular issue (e.g., discriminant validity testing) is that it is likely reduce the overall power of the analysis. In other words, it leads to more false negatives.
For example, one could simply recommend using the AVEs on the diagonal of the table traditionally used for discriminant validity, instead of the square roots of the AVEs. This would lead to a more stringent test, but would also lead to more false negatives.
We conduct extensive Monte Carlo simulations before adding features into WarpPLS, and leave out features that could lead to misleading conclusions. WarpPLS could have many more features than it currently has.
I suggest users employ the full collinearity VIFs, which can also be also as measures of common method bias. In a presentation in the recent PLS Conference in Macau, Dr. Mostafa Rasoolimanesh demonstrated that the criterion that full collinearity VIFs be lower than 3.3 does a generally better job than HTMT in tests of discriminant validity.
I recommend that this be done in an integrated way, with other measurement quality assessment tests. For example, discriminant validity violations can also be identified through the combined loadings and cross-loadings table provided by WarpPLS, where loadings are unrotated and cross-loadings oblique-rotated.
Discriminant validity violations are often associated with oblique-rotated cross-loadings greater than 0.5, and with full collinearity VIFs greater than 3.3. For the foundation of these full collinearity VIFs, see the following publications, which are available in full text from warppls.com.
Kock, N. (2015). Common method bias in PLS-SEM: A full collinearity assessment approach. International Journal of e-Collaboration, 11(4), 1-10.
Kock, N., & Lynn, G.S. (2012). Lateral collinearity and misleading results in variance-based SEM: An illustration and recommendations. Journal of the Association for Information Systems, 13(7), 546-580.
Ned
--
Join us at the PLS Applications Symposium: http://plsas.net
---
You received this message because you are subscribed to the Google Groups "PLS-SEM" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+unsubscribe@googlegroups.com.
To post to this group, send email to pls...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Florian
Dear Ned,
Could you please elaborate a little bit more on your following statement:
The problem with any new method that is aimed specifically at a particular issue (e.g., discriminant validity testing) is that it is likely reduce the overall power of the analysis. In other words, it leads to more false negatives.
What is the overall power of an analysis? And why is it reduced by using new methods that aim at a particular issue?
Thank you very much in advance and best regards,
Florian
--
To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+u...@googlegroups.com.
To post to this group, send email to pls...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
--
Join us at the PLS Applications Symposium: http://plsas.net
---
You received this message because you are subscribed to the Google Groups "PLS-SEM" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+u...@googlegroups.com.
Ned Kock
Hi Florian.
New tests and related criteria can be devised to pick up 100 percent of violations of arguably any measurement quality issue. The tendency of course is for methodological researchers to come up with increasingly more stringent tests.
The problem is that users of these new tests and related criteria would often incorrectly conclude that their measurement models do not pass those criteria. They would thus assume that their structural model results cannot be trusted, when in fact that is not the case.
In doing so, they would be committing false negatives, and thus compromising the overall power of their analyses. Broadly speaking: power = 1 – (probability of committing a false negative).
The bottom line is that any new test and related criteria devised have to be considered in a broad context; e.g., what is its relationship with other tests that pick up related problems, such as the full collinearity assessment tests implemented in WarpPLS? Also, the effect of the new test and criteria on power has to be carefully considered.
Ned
Jose L. Roldan
__________________________________________________________ Dr. José L. Roldán Professor of Management Senior Editor, The DATA BASE for Advances in Information Systems http://sigmis.org/the-data-base/ Department of Business Administration and Marketing Universidad de Sevilla Ramon y Cajal, 1. 41018 Seville (SPAIN) Voice: (34) 954 554 458 / 575 Fax: (34) 954 556 989 Skype: jlroldan67 <mailto:jlro...@us.es> URL: http://personal.us.es/jlroldan
http://orcid.org/0000-0003-4053-7526
Google Scholar: https://goo.gl/PPY32K __________________________________________________________