Matt Keough posted on Friday, February 13, 2015 - 2:32 pm
I am running a model with 4 predictors (continuous), 2 mediators (continuous), and one dichotomous outcome. I ran a model with direct effects from 2 of the predictors and the model fit the data well. Weirdly; however, when I test full mediation (without the direct effects), my model doesn't converge. I tried increasing the iterations, but I get an error stating that there is a linear dependency between the two mediators. This is confirmed when I check the Tech 4 output. But at the bivariate level, the correlation between the two mediators is only .37. Any idea what might be causing the issue in the mediation model minus those two direct effects?