med_smean: Structural means model to compute controlled direct effects...

In kaskarn/causamed: Mediation in the causal inference framework

Description

med_smean

Usage

med_smean(dat, A, Y, M, C = NULL, L = NULL, boot = 10, nmin = 10,
  mlvl = NULL, quants = c(0.025, 0.5, 0.975), mids = NULL, maxit = 5)

Arguments

dat	a dataframe containing the exposure, outcome, mediators, and confounders
A	the exposure of interest. Can take any form
Y	the outcome, currently must be continuous
M	the mediators of interest
C	confounders of either X -> M and/or M -> Y. Can take any form
L	the exposure-induced confounders of the association of M with Y
boot	the number of bootstrap samples used to build confidence intervals
nmin	number of participants all categories of exposure must have; samples will be redrawn if this criterion is not met
mlvl	the levels of M to calculate corresponding CDE's to. Default is sample average.
quants	an optional vector of quantiles for the confidence interval (95 percent by default)
mids	an optional mids object to serve as template for imputations

Details

Returns the controlled direct effect CDE(M) of an exposure A on an outcome Y, not operating through a mediator M, in the case of exposure-induced confounding of the M->Y association. Follows section 5.3.5 of Vanderweele's book on causal mediation analysis.

Value

An S3 object of class cmed_smean containing:

k2 the coefficient of M in regression of A, M, L and C on Y

k3 the coefficient of A*M from regression of A, M, L and C on Y

g1 the coefficient g1 from the regression of A and C on partial residuals of Y

ymod1 the model of Y for the last bootstrap sample

ymod2 the model of Y residuals for the last bootstrap sample

cde array story controlled direct effects from bootstrapped results

Examples

my_list <- med_smean(data, A, Y, M, c1 + c2 + c3 + c2*c3, mlvl = quantiles(M, probs = c(0.25, 0.5, 0.75)))

kaskarn/causamed documentation built on Dec. 28, 2021, 11:01 a.m.