Hi,

I have a question regarding power to detect effects when probing an interaction. Let say I probe a significant interaction (continues predictor, moderator and outcome) at 1 SD above or below the mean. My question is, does the actual number of cases that fall  above/below 1SD impact the power to detect effects? I know it is absurd to probe the interaction in imaginary values of the moderator, but not sure whether it affects the statistical power. Any thoughts?

Similar questions and discussions