forum.smartpls.com

I am trying to do a "consistent PLS Algorithm" and the result is erroneous with N/A present in my model and matrix. Please take a look at the attached pictures and advise.

This kind of outcome occurs when the PLSc manipulation of PLS results entails extreme values. The reason are bad data and/or poor model fit.

Best
Christian

See also viewtopic.php?f=5&t=3531

In this thread and the one that jmbecker has referenced (viewtopic.php?f=5&t=3531), this problem is attributed to the consistent PLS algorithm in reflective measurement.

I encounter this problem also with a formative construct and the standard PLS algorithm. Are the underlying reasons the same?

Thank you.

In such a model it is more likely the case that you have a severe multicollinearity problem or some other data issues that cause the problem.

Thank you, jmbecker. I don't think that it's a (multi)collinearity problem since all VIFs are between 1.19 and 1.70.

If I include the same construct (INFO with INFO1-8) in a larger model with other constructs, this problem doesn't occur.

You may send your project (including model and datafile) to our support and they can have a look at the issue.

Thank you for sending us your request.
The problem is that you only have one construct in your model. Such a model is not estimable with PLS. The normal PLS algorithm needs at least two connected constructs for identification of the model.
It seems to be a small bug in the latest release that the path model does not invalidate. The specific validator seems to be inactive. Previous releases would not have allowed such a model to turn valid and thus would prevent users from estimating algorithms on such an invalid model. We will fix that in the next release.

Thank you very much for looking at the issue and explaining the problem - much appreciated!