Message/Author 

Celeste. C. posted on Thursday, July 26, 2012  2:45 am



I have a simple mediation model, i would like to know if there is any thumb rule to claim that a direct/ indirect effect (standardized regression weight in SEM) is of practical importance? Appreciate if there is any reference. Thank you. 


If the covariate is binary, you can use the values for Cohen's effect size. 


Thanks Dr Muthen. How about indirect effect size? Any thumb rule for tha? 


It's the same. If the observed exogenous variable in the indirect effect is binary, use Cohen's effect size. 

K C posted on Saturday, March 19, 2016  7:37 pm



Hello there, 1. What if the mediator is not binary (continuous)? Can the standardized indirect effect coefficients still be interpreted as "sizes" of the effect, as per Cohen? 2. I am trying to determine the effect sizes of a mediation model that goes through several paths (M1 > M2 > M3). Is the coefficient interpreted the same way or would it be different since there is additional multiplication of the indirect paths, leading to smaller coefficient values? Thanks! 


1. Yes. The scale of the mediator does not matter. 2. The size of the indirect effect need only be considered with respect to the SD of the x variable (the exposure) and the y variable (the outcome). 

Back to top 