I know that we can test mediation models in SEM. In full mediation, we see that the c' path becomes insignificant. But in the case of the partial mediation we need another index for the significance of the decrease in the coefficient. Is it possible to use Sobel Z or bootstrapping? If the answer is yes, then which output values/data should I use? References to useful instructions and examples are very much appreciated. I use LISREL 8.51 software.