Interpretation: significant effects, but low f2 and q2

[Victoria01]

Hi everyone,

I have multiple significant effects, but some of the paths exhibit very low f2 and q2 values. Does that mean that for these paths, even if the significance of an effect would support my hypothesis, I would have to reject a hypothesis?

Similarly, if I wanted to compare multiple groups via MGA: if some of my groups' constructs had very low R2 and Q2 values but I still observed significant differences, how would you interpret the relationships? Does that also mean that despite significant differences, one cannot derive any conclusions because explanatory power and predictive relevance are too low?

Thanks!

[jmbecker]

With the end of second statement you are heading in the right direction.
First, there is a difference between significant and substantial. You might have significant effects (i.e., significantly different from zero), but still not substantially large to make a difference in practical applications. As sample size increases even very small effects become significant. Yet, the predictor might not contribute much additional explanation / predictive power to the model, because your other variables already explain most of the variation in your dependent variable. Thus, the variable has an effect, but it is very small.
Second, you observe generally very low R² and Q² values. That means that your model has low explanatory and predictive power. However, low is always judged relative to what you expect. If you are in a customer satisfaction context, where the predictors of satisfaction are well known, you want a model that has high R² and Q² values. However, if you are in a field that researches a very new phenomenon that might be hard to influence (e.g., some consumer behavior based on a subtle treatment effect) then you might expect low R². In this case it is more important that you even find an effect or that there are differences across groups.
Hence, it also always depends on the context and what you expect.