decomp: Causal Effect Decomposition Based on a Fitted 'rwrmed' Model

In xiangzhou09/rwrmed: Regression-with-residuals (RWR) Analysis of Causal Mediation

Description

decomp is a function that implements causal effect decomposition based on a fitted rwrmed model. It returns a two-component decomposition of the total effect into the randomized interventional analogues of the natural direct effect (rNDE) and the natural indirect effect (rNIE). It also returns a four-component decomposition of the total effect into the controlled direct effect (CDE) and the randomized analogues of the reference interaction effect (rINTREF), the mediated interaction effect (rINTMED), and the pure indirect effect (rPIE).

Usage

decomp(object, a0 = 0, a1 = 1, m = 0, bootstrap = TRUE,
  rep = 250)

Arguments

object	An object of class rwrmed.
a0	The baseline level of treatment.
a1	The level of treatment to be contrasted with the baseline.
m	The level of the mediator at which the CDE is evaluated.
bootstrap	Whether to compute standard errors and 95 nonparametric bootstrap.
rep	Number of bootstrap replications if bootstrap = TRUE. Default is 250.

Value

A list of two elements.

twocomp	Two component decomposition of the rATE into rNDE and rNIE.
fourcomp	Four component decomposition of the rATE into CDE, rINTREF, rPIE, and rINTMED.

See Also

rwrmed for implementing the regression-with-residuals (RWR) approach to causal mediation.

a0 = 0; a1 = 1; m = 0; bootstrap = TRUE; rep = 250

xiangzhou09/rwrmed documentation built on May 27, 2019, 2:08 p.m.