Dear MLwiN Forum,
I am conducting simple multilevel mediation. That is, I have an IV, MV and a DV. The model I have specified is a random intercept fixed slopes (based on loglikelihood ratio tests). I have conducted the Baron & Kenny approach together with a Sobel test of the indirect effect (given that it is a fixed slope model there is no need to add the covariance of the slopes to ab). However, on further reading I believe that there are problems with Baron and Kenny’s approach and the use of the Sobel test. The received wisdom from the reading I have done is that a boostrapping method to test the significance of the indirect effect is perhaps the most robust method. However, I am not sure if MLwiN can perform such an analyses? if so how? if not does anyone have any recopmendations?
Many thanks,
Calum