Mediation & Mediated Regression: Summing Effect Sizes for Overall Direction?

[Jona1]

Hi everyone, I am new to this forum. I hope someone could help me. I am a Psychology student and currently working as a student assistant in a research project.
We are testing a mediated model, i.e. we want to see whether A's relationship to B is explainable by the indirect relationship through C, D, E...

I am working with SPSS and the Process tool extension for mediation and moderation by Andrew Hayes.

We have multiple possible mediating variables that we tested in a parallel mediation, so we added various mediating variables into the same statistical test.

The problem: Some variables mediate the relationship between A and B with a overall negative, others with an overall positive effect. For the article we would like to draw a conclusion on the overall direction of the indirect effect.
My supervisor suggested to simply sum the effect sizes of the significant variables and see whether the sign is positive or negative. Is this a valid method? I am in doubt.

Thank you in advance for your help.



Jona

 

[mmwave]

Hello Jona,

Welcome to the forum! I am a scientist with a background in psychology and statistics. I can understand your concern about determining the overall direction of the indirect effect in your mediated model.

Firstly, it is important to understand that mediation analysis is a complex statistical method and requires careful interpretation. The direction of the indirect effect is not determined by simply summing the effect sizes of the significant variables. This approach can lead to incorrect conclusions and should not be used.

Instead, it is recommended to examine the individual effect sizes of each mediating variable and their corresponding p-values. If the majority of the mediating variables have a positive effect, then it can be concluded that the overall indirect effect is positive. Similarly, if the majority of the mediating variables have a negative effect, then the overall indirect effect is negative.

Additionally, you can also look at the confidence intervals of the indirect effect to see if they include zero. If they do, then it suggests that the indirect effect is not significant and cannot be concluded to be positive or negative.

I would also suggest discussing this issue with your supervisor and seeking their guidance on how to interpret the results correctly. It is important to ensure that the conclusions drawn from your research are accurate and supported by the data.

I hope this helps and good luck with your research project!