I have noticed that while running three mediation analyses the IV to M path produces the same coefficient in all three of my mediation analyses. The DV in these analyses are categorical. The IV and M are the same across all analyses. I am using bootstrapping to calculate confidence intervals for each of the effects, but find that the IV and M effect for each analysis produces different confidence intervals, even though the relationship between the IV and Mediator is the same (e.g. .482 in each case). I wondered whether this was because the confidence intervals are bootstrapped and for each analysis new samples are drawn that then produce differences between Confidence intervals.
So the questions are: 1. Is the above the reason for the differences between CIs? 2. In order to avoid this, should one set up the analysis in such a way that all DVs are tested simultaneously and only one bootstrap procedure is used, rather than running three separate analyses? If so, what would this look like in the syntax?
The code I am using for a single mediation analysis is this: VARIABLE: NAMES ARE DV1 DV2 DV3 IV M; USEVARIABLES ARE DV1 IV M; CATEGORICAL ARE DV1; ANALYSIS: bootstrap = 5000; MODEL: M on IV; DV1 on M IV; MODEL INDIRECT: DV1 via M IV; OUTPUT: cinterval(bcbootstrap);