Mediation (indirect effect)

A mediation asks whether X affects Y not only directly but also through an intermediate step M (the mediator). Three regressions give the paths: a (X → M), b (M → Y holding X constant), c′ (X → Y direct) and c (X → Y total). The indirect effect is the product a·b; c = c′ + a·b holds. If the bootstrap confidence interval of a·b excludes 0, this is statistical evidence for an indirect effect (it does not establish a cause). The proportion a·b / c shows which part of the total effect runs through M.

Example: ab = 0.42, 95% bootstrap CI [0.27, 0.58]: the indirect effect through self-efficacy is significant, and so is the direct effect (c′ = 0.42). This is consistent with support also acting on well-being through self-efficacy - a single survey cannot prove it.

Caution: With data from a single survey, a mediation does not establish causality: the order X → M → Y is an assumption from theory. Whether a and b are individually significant does not decide on the indirect effect; the interval of a·b does.