I am trying to figure out how to interpret my data where I think I have inconsistent mediation (MacKinnon, Fairchild & Fritz, 2007).
Before including covariates:
I have a non-significant total effect of IV on DV however significant effects of IV on M and M on DV. I still have a significant but reduced direct effect of IV on DV, which I have read can still mean there is mediation (Valeri & VanderWeele, 2013).
After Including covariates:
I have a significant total effect of IV on DV and significant effects of IV on M and M on DV. I now have a non-significant direct effect of IV on DV.
Best Answer
Without the covariates you had a partially mediated model ie IV was significantly related to DV and IV to M and M to DV were significant. However, the overall model was not significant. With the full inclusion of the covariates you now have a more robust model; the ability to carry the CV load and remain a significant total effects model is testament to that. This model is fully mediated model, ie the IV to DV relationship is non-significant, while the IV to M and M to DV are significant (did not see the mediating data in your figures). The latter is a stronger model.