I have been running some models with two mediating variables (both latents) and have come across some somewhat strange findings I am not sure how to interpret.
In a nutshell, we are finding that the total indirect effect is significant, but the specific indirect effects are either marginally significant or not significant.
For example, we are predicting child PTSD symptoms from parental IPV, looking at both parenting and psychopathology as mediators. In the model, we correlate parenting and psychopathology, and examine both mediators simultaneously.
The total indirect effect is significant, parenting is only marginally significant (p=.07), and psychopathology is not. The correlation between parenting and psychopathology is .27 when we use IPV perpetration in the model, and .39 when we use IPV victimization in the model.
So how can we interpret this? Do we just say it is clear that both parenting and psychopathology exert an influence on child PTSD symptoms, but we can't specify how each independently effects child PTSD symptoms? Does that also mean that because parenting and psychopathology are moderately correlated that we can't really "tease" apart their independent influence?
I'm making some mistake I can't figure out in using the MODEL INDIRECT command. Here is my model:
MODEL: ! all vars latent, MLR outcome ON coping dist_3d readmitb ; dist_3d ON coping ; MODEL INDIRECT: outcome IND dist_3d coping ;
Output says, "TOTAL, TOTAL INDIRECT, SPECIFIC INDIRECT, AND DIRECT EFFECTS", but I'm only getting the specific indirect effect in the output, and I'd like the total effect also. What am I not specifying correctly?
How do I interpret output where there is a sig effect from x to y, m to y, and x to m however the total effect is non sig whilst the indirect and direct effects are sig? Does this refer to inconsistent mediation and does this mean that it is partially mediated?