forum.smartpls.com

Hi all,

I have a model with:
1 IV
2 Mediators (serial mediation)
2 DVs
1 Moderator (between direct link IV --> DV1 and IV --> DV2)

First question about effect size: I calculated the f2 value with SmartPLS and I have found six relationships with no effect, namely: Moderator link 1, moderator link 2, mediator 1--> DV1, IV --> DV1, IV --> DV2, and IV --> Mediator 1

My question now is, what does this mean? I follow the logic of Cohen saying that 0.02, 0.15, and 0.35, respectively represent small, medium, and large effects of the exogenous variable. But how do I write about this in my thesis and does it have consequences? Important to note: the relationships that have no effect are also non-significant.

Second question about Q2.

I have determined that my Q2 values are above 0, so the predictive relevance is guaranteed. Is it necessary or recomended to also calculate q2?


Thank you very much in advance!

I attached a picture of my model.

Anyone has an idea perhaps?

First:
It does mean that there is no effect. At least on average (if you are looking at the awareness effects, but have a significant interaction effect).
Btw, I would usually report it the other way: first significance (if not significant than there is no effect), second strength (i.e., effect size f²) to answer wether relation is substantial when it is significant.

Second:
That depends. I personally do not like the q² effect sizes so much, because there is a lot of controversy about how to correctly calculate them, but generally it could be interesting to look at them too.

Dear,
please for one explanation.
I know that if my Qsquare of the endogenous construct is higher than zero that I have predictive relevance. But how do I interpret the strengths i.e. week, medium, and strong?
In the material of the workshop it is written 0.02, 0.15, and 0.35 but in the articles Hair et al (2019); When to use and how to report the results of PLS-SEM and Hair et al (2019): Assessing measurement model quality in PLS-SEM using CCA, it is written 0, 0.25 and 0.50.
Is it a mistake in the workshop materials or can I use that ranges for Qsquare? The picture in attachment.
Kind regards,

First, thresholds are never right or wrong. They are always arbitrary and to some degree based on the opinions of the authors or common practice in the field.

The thresholds in the course slides that you show are based on the PLS Primer book 2nd edition. However, we do not use this slide anymore in our courses because we do not suggest to use Blindfolding anymore. Most current recommendations in the PLS literature suggest using PLSPredict instead of Blindfolding for assessing predictive relevance:
Shmueli, G., Sarstedt, M., Hair, J.F., Cheah, J.-H., Ting, H., Vaithilingam, S., & Ringle, C.M. (2019). Predictive model assessment in PLS-SEM: Guidelines of using PLSpredict, European Journal of Marketing, 53(11), 2322-2347.

Some researchers still like to report blindfolding Q² to facilitate comparability with older research. But I would certainly complement this with an analysis of PLSPredict.

Dear Mr. Becker,

thank you very much for the answer. I read the mentioned article and made the PLSpredict analysis. The results show high out-of-sample prediction.

Kind regards,
Antonija