How to quantify the effect of a mediator, SPSS, Psychology experiment I have collected data to answer the question “Does Neuroticism Mediate the effect of a Mindfulness Intervention?” I Have collected pre-intervention Mindfulness Score and a post intervention measure (both scale). I also took a neuroticism score before the intervention to estimate the level of trait neuroticism (also scale). There was no control group.
So far I have managed to demonstrate with a paired-samples t-test that there is a significant difference between pre and post-mindfulness scores, I have used Pearson correlation to show that mindfulness and neuroticism are negatively correlated at baseline (both of these are a replication of the existing findings).
For the mediator part, I intended initially to use the model from another study which divided the neuroticism score into low, low/medium, medium high and high as a quasi-independent variable. However, the scores in my data were too clustered together without much of a range in the group (there was a high attrition rate ~45%) so these 4 groups are not possible. I have tried running a repeated measures ANOVA, where I dichotomize neuroticism into low/high, which I know is not the best idea as I will artificially restrict the variance in my data. This does however produce quite a sensible looking plot and desirable output for my study but I know it's not sound practice.
I have also tried doing what I think is a repeated measures ANCOVA, with time at 2 levels as a within subjects factor (pre and post-intervention score) and neuroticism as a covariate but I have nothing to put into the between subjects factor so I’m not sure if it’s accurate. It is giving neuroticism a value of 3.45 in the model but I have no idea what that number means or how to interpret it. It also produces a sensible looking plot showing the increase in mindfulness pre to post-intervention. I have also tried is computing a mindfulness score difference and correlating that with the level of mindfulness which also gives me a medium size positive correlation <0.05.
Any help would be greatly appreciated as I am really desperate!

 A: You cannot answer a mediation question with this design, but I'm not sure you want to. It seems like you are more interested in moderation, i.e., does baseline neuroticism affect how effective the intervention is? You can define "how effective the intervention is" as the mindfulness difference score between post and pre. So the question is whether there is an association between baseline neuroticism and this difference score. This is a simple regression or correlation analysis, and there is no need to dichotomize any variables. You can just produce a scatterplot of the difference scores on the y-axis and neuroticism on the x-axis and see how related the variables are using correlation or fit a regression line through the points.
Know, though, that the design of this study is very limited and does not allow you to make any causal conclusions without strong assumptions. I doubt it would be publishable as-is. First, without a control group, you can't say whether the intervention was effective or not; it's entirely possible participants' mindfulness scores would have changed in the same direction in the absence of the intervention, e.g., simply because of the structured time or because of external factors (i.e., maturation). Second, even if you show that neuroticism is associated with changes in the outcome, that doesn't mean neuroticism itself causes those changes; it could be that neuroticism is correlated with a different variable that is actually responsible for those changes. As a pilot study or a class project this is fine; as a study to unequivocally demonstrate the effectiveness and mechanism of an intervention, this is inadequate.
