Serial mediation in path analysis PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Sabahat Cigdem Bagci posted on Friday, January 12, 2018 - 12:59 am
Hello, I have a path model with 2 consecutive mediators. So I have such a partially mediated model:
X->M1->M2->Y
When I look at model results, Y is not significantly predicted by X, M1, or M2.
But when I look at indirect effects with 95% CI intervals, it seems like the indirect paths of X->M1->Y and X->M1->M2->Y are significant. What may be the reason? Should I assume that the mediation is significant even though the mediators do not predict the outcome variable?
Thanks
 Bengt O. Muthen posted on Friday, January 12, 2018 - 5:10 pm
Try bootstrapping and look at these confidence intervals instead. See e.g. the Table 2.21 input shown for our book at

http://www.statmodel.com/mplusbook/chapter2.shtml

Also, perhaps your mediation model does not fit well so that the estimates are not meaningful.
 Sabahat Cigdem Bagci posted on Monday, January 15, 2018 - 10:35 pm
Hello, thanks for the answer. The fit of the model is good. So when I look at bootstrapped results for the direct effects, it seems like I have a significant path from the mediator to the outcome. Is it ok to report the CI for the significance of direct effects? When I look at papers, I usually see point estimates and significance level for direct effects, and CI's used for the indirect effects.
 Bengt O. Muthen posted on Tuesday, January 16, 2018 - 2:29 pm
CIs can always be used, so also for direct effects.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: