I'm having trouble with the mediate() and medsens() functions from the mediation package. My treatment, mediator and outcomes variables are all continuous and scaled to the 0-1 interval. The data is observational not experimental. I am thus using lm() for the initial outcome and mediation regression models. I also use weights as the data are survey data. (As a footnote, I use the svyglm() function from the survey package in other analyses, but svyglm() models are not compatible with mediate().) The help file for plot.medsens() reports that "When rho is zero, sequantial ignorability holds, so the estimated value at that point will be equal to the estimate returned by the mediate. The confidence level is determined by the ?conf.level? value of the original mediate object." This is not the case with my data. The value of delta^1 or delta^0 when rho = 0 is not the same as the Mediation Effect produced by the mediate() function. Through trial and error I have pinned the cause down to my use of weights in the original lm() models. I'm not sure why the value of the mediation effect should differ when across mediate() and medsens() when using weights. Is there something I am doing wrong, or missing? many thanks for any advice / help. Chris