# Trying to understand mediation analysis

#### char1

##### New Member
Hi,

I am currently completing a mediation analysis for the first time using PROCESS in SPSS. The model is as outlined below.

Model 1
Predictor: Disordered Eating (DE)
Mediator: Stigma
Outcome: Depressive symptoms (DS)

Model 1
Predictor: Disordered Eating
Mediator: Self-compassion
Outcome: Depressive symptoms

I found that:
1. DE predicts stigma and self-compassion
2. Stigma and self-compassion predicts DS
3. Stigma and self-compassion do not mediate the relationship between DE and DS.

I am trying to come up with reasons why they did not mediate the relationship but am getting myself confused with what reasons there could be?

Essentially I am not sure what would mean that the separate regressions (point 1 and 2 above) are sig but the mediation is not. Does it essentially mean that stigma and self-compassion might be important to the two variables separately but does not explain the relationship between the two?

Hope this makes sense

Last edited:

#### Karabiner

##### TS Contributor
How can you tell from the results that a predicts b, or that something does not mediate?

With kind regards

Karabiner

#### char1

##### New Member
How can you tell from the results that a predicts b, or that something does not mediate?

With kind regards

Karabiner
I did a simple linear regressions
1. DE predicting Stigma > beta was P<.001
2. Stigma predicting Depression > beta was P<.001

I then used PROCESS to run a mediation analysis and the Confidence intervals in the indirect models (one for stigma as mediator and one for self-compassion as mediator) both contained 0 meaning it was not a significant mediation?

#### Karabiner

##### TS Contributor
Please, do not report just p-values or whether a confidence interval contains 0.
These are mostly useless information, at least if you do not report also the size of the coefficients
and sample size. And you did not mention how DE and depression were related.

With kind regards

Karabiner

#### char1

##### New Member
Sorry I was planning on reporting the size of the coefficients with that and yes DE was able to predict depression.