ZYmediate function

Robust mediation test

Performs a robust mediation test according to the method proposed by Zu & Yuan (2010).

ZYmediate(x, y, med, nboot = 2000, alpha = 0.05, kappa = 0.05, ...)

Arguments

• x: vector with predictor values.
• y: vector with response values.
• med: vector with mediator values.
• nboot: number of bootstrap samples.
• alpha: alpha level.
• kappa: the percent of cases to be controlled when robust method is used.
• ...: currently ignored.

Returns

ZYmediate returns an object of class "robmed" containing:

• a.est: effect of predictor on mediator)

• b.est: effect of mediator on response (in multiple regression model which includes the predictor as well)

• ab.est: indirect effect (mediation effect)

• CI.ab: confidence interval mediation effect

• p.value: p-value mediation test

• call: function call

References

Zu J., Yuan, K.-H. (2010). Local influence and robust procedures for mediation analysis. Multivariate Behavioral Research, 45, 1-44.

Examples

## Leerkes data:
## Y: Efficacy
## X: MatCare
## M: Esteem

## fitting robust mediator regressions
require(MASS)
summary(rlm(Efficacy ~ MatCare, data = Leerkes))
summary(rlm(Esteem ~ MatCare, data = Leerkes))
summary(rlm(Efficacy ~ MatCare + Esteem, data = Leerkes))

## robust testing of mediating effect (indirect effect)
with(Leerkes, ZYmediate(MatCare, Efficacy, Esteem))