med_reg: 'med_reg' Gives controlled direct effect, natural indirect...

In kaskarn/causamed: Mediation in the causal inference framework

Description

med_reg Gives controlled direct effect, natural indirect effect, natural direct effect using regression-based formulas affording the best statistical power. The procedure is described in chapter 3.1.1 of Tyler's book

Usage

med_reg(df, X, M, Y, C = "", noint = FALSE, mlvl = NULL, delta = FALSE,
  ref = 0, treat = 1)

Arguments

df	Data frame
X	exposure, can be continuous or categorical
M	mediator, must be continuous for the moment
Y	outcome, must be continuous for the moment
C	confounders of exposure-outcome or mediator-outcome association. Can take any form
noint	supress exposure-mediator interaction term
mlvl	vector of values at which to compute a controlled direct effect
delta	whether the delta method should be used to get confidence intervals
ref	referent level of exposure, 0 by default
treat	treatment level of exposure, 1 by default

Value

a list composed of the natural direct effect, natural indirect effect, total effect, and a controlled direct effect for each line given in mlvl

Examples

my_list <-med_reg(my_df,
 X = "my_exposure",
 M = "my_continuous_mediator",
 Y = "my_continuous_outcome",
 C = c("confounder_1", "confounder_2"),
 treat = high_exposure_level,
 ref = low_exposure_level,
 mlvl = seq(25, 40, by=5) ## possible BMI values for the mediator

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