Effect Sizes and Total Effects

[Sam]

Hey Guys,

I have a question concerning effect sizes and total effects, I hope you can help me.
My model has 1 exogenous, 1 endogenous and 3 mediating variables. A direct path from exogenous to endogenous variable exists. Does the reported effect size f² consider only the direct connection from exogenous to endogenous or does it even consider the indirect effects? In other words: is the effects size of the direct effect different from the effect size of the total effect? If yes, is there a way to calculate effect size f² for the total effect?

i hope you can understand what i mean :D

Thanks in advance!
Sam

[jmbecker]

It only considers the direct effect. What the effect size f² calculates is the change in R² if you exclude a variable from the model. If you exclude your exogenous variable, the mediators will stay in the model and will still effect the final dependent variable, hence only the direct effect of the exogenous is excluded.

I'm not sure if I understand your second question. The effect size of the total effect in your model seems to be the R² itself, because if you exclude all effects of the exogenous variable then there would be no variable left in your model and hence R²=0. (or do you have other variables predicting the endogenous variable that are not involved in the mediation model?)

The effect size for an indirect effect via a single mediator has to be calculated by yourself (excluding the exogenous and focal mediator from your model and then calculating f²). Be aware that we calculate f² with the final LV scores fixed. If you simple exclude variables from your model, the weights and hence LV scores may change and may give you false results.

[Sam]

Dear Dr. Becker,

thanks a lot for your explanation. I painted my original research model (its a bit more complex than i wrote in my first post), it looks like this:

researchmodel.png (25.86 KiB) Viewed 5608 times

I hope my second question gets clearer. For my research the total effects of the variables 8, 9 and 10 on variable 3 are relevant. I understand now, that the f² given by smartpls are just those f² of the direct effects. What I try to do now is to calculate the f² of each total effect of the variables 8, 9 and 10 each incl. all indirect effects. Is that possible without any distortions?

Thanks again for any hints!
Sam