Hi all, Is it possible to test mediation effects with a censored outcome variable using the 'mediation' package? While on p. 1f. of Imai, Keele, Tingley and Yamamoto (2011) the tobit model is mentioned for use with 'mediate', Table 1 (p. 8) shows that tobit via vglm cannot be used to estimate causal mediation effects. Also help(mediate) does not mention 'vglm' class objects as possible arguments of 'mediate'. So, is calculating mediation effects using 'mediate' with vglm-tobit models possible at all? Thanks in advance, Niklaus --- Imai, K., Keele, L., Tingley, D., & Yamamoto, T. (2011). Causal mediation analysis using R. http://www.polisci.ohio-state.edu/faculty/lkeele/mediationIII.pdf