Effect Size - f-squared in PLS3

[Charly2309]

I have read a lot of articles regarding calculation of f-squared, however I still face some problems and have difficulties to interpret my results.

1) Did I understand it right, that SmartPLS 3 is calculating f-squared automatically, so I do not have to eliminate variables to calculate the f2?
2) I attached my basic model and the results I obtained when I calculated the f-squared. Could someone tell me what I can read from these numbers?

Bildschirmfoto 2016-11-26 um 13.42.57.png (235.99 KiB) Viewed 6757 times

Bildschirmfoto 2016-11-26 um 13.42.24.png (71.64 KiB) Viewed 6757 times

What can I interpret from the values when considering 0.02 --> weak, 0.15 --> moderate and 0.35 --> substantial?

Thanks a lot for any help!!! :-)

[jmbecker]

1) Yes.

2) The f² effect size is a standardized measure of effect size. The numerator of the f² reflects the proportion of variance uniquely accounted for by the focal variable, over and above that of all other variables in the regression. It is set relative to the unexplained variance (i.e., 1 minus the variance explained by all variables in the regression).

In my view the last part is often the point of confusion and makes this effect size measure a little bit unintuitive at first glance. The upper bound is not 1 (as one might expect), but infinite.
The effect size measure becomes larger if either 1) the focal predictor explains more variance relative to the other predictors or 2) more variance is explained in total.

If your focal variable explains 20% additional variance and all other variables explain no variance (numerator is 0.2 and denominator is 1-0.2) you get an f² of 0.25.
If your focal variable explains 20% additional variance and all other variables explain 60% of variance (numerator is still 0.2 and denominator is 1-0.8) you get an f² of 1.