When discussing topic of Mediation, Kenny concluded several specification errors including reverse causal effects, measurement error in the mediator, omitted variables. There are also some extensions such as latent variables, covariates, dichotomous variables and so on. Kenny also recommended other kinds of test such as Sobel test (http://www.people.ku.edu/~preacher/sobel/sobel.htm) and bootstrapping by using either SPSS (http://www.people.ku.edu/~preacher/) or Amos (http://amosdevelopment.com/video/indirect/flash/indirect.html) | 
超级强大
