I want to do a mediation analysis for variables Y,X,M in statistical package R.
I have perform the following regression models:
Y=intercept1+a0*X
M=intercept2+a*X
Y=intercept3+c*X+b*M
For the mediation analysis i have used the package: library(mediation)
The results from the estimated parameters from regression models are:
a0=-0.86
a =-1.01
c =-0.19
b = 0.66
The parameter c is non significant (p=0.2).
The results from mediation analysis are:
Estimate 95% CI Lower 95% CI Upper p-value
ACME -0.660 -0.878 -0.45 <2e-16 ***
ADE -0.195 -0.503 0.07 0.2
Total Effect -0.855 -1.178 -0.52 <2e-16 ***
Prop. Mediated 0.772 0.535 1.12 <2e-16 ***
I don't know how to interpret the negative but statistically significant ACME (a0-c) parameter. Can someone help me ?