interpretation mediation


I'm learning to interpret mediation analysis, but I'm not really getting it. I have the following "problem":

Mediation 1: Since the direct effect between x1 and y is not significant, but the paths through M are, can I conclude that this is a full mediation?
Mediation 2: I don't understand this one at all.. how can the path between m and y be significant when the other 2 aren't?
Mediation 3: Can i conclude that M isn't a mediator since the indirect effect of X is nog significant?

I calculates all these path's and p-values, so I hope i did this correct. If it isn't, then that is a lesson for next time.



Less is more. Stay pure. Stay poor.
I would not focus so much on pvalues which could be influenced by sample sizes and they neglect to convey magnitude. I would provide estimates with SEs.

Are these real data or just toys you are playing around with?

Strong context knowledge is needed for mediation analysis, it can't just be used as an exploratory analysis and of note you can have fully mediated and partially mediated effects. So M1 is closer to fully mediated. Not an expert in this area, but it is always interesting to remove the mediator and see what happens. Are you familiar with Markov independence - it would say that the future is independent of the past given the present. So if you wanted to know the genetics of a grandbaby, knowing grandparents DNA wouldn't provide anything beyond what the parents' DNA could provide. This is an example of direct mediation. When knowing the variable of interest and mediator (both) helps understand the DV then you have partial mediation. Yeah, in M2, X2 seems independent of M and Y, but I would be curious its affect with M removed. Are you using something like the PROCESS to get the above estimates?

Final comment, the back ground knowledge of the variables is needed to support the use of the models.