iorw: Inverse odds ratio weighting.

In adayim/causalMed: Causal Mediation Analysis in Survival settings

Description

Mediation analysis for time fixed mediator using inverse odds ratio. Output contains total effect, natrual direct effect and natural indirect effect. The confidence interval will be calculated using 1000 bootstrap with normal approximation.

Usage

iorw(
  fitY,
  data = NULL,
  exposure,
  mediator,
  family,
  ref = NULL,
  stabilized = TRUE,
  R = 1000
)

Arguments

fitY	model object of the final outcome, all variables of interest should be included except mediators.
data	Data set to be sued, if NULL, the data from outcome model will be used.
exposure	Intervention/Exposure variable, control or unexposure must be set to 0.
mediator	Name of the mediator(s).
family	a description of the error distribution and link function to be used in the exposure model. Only binomial, multinomial and gaussian supported now.
ref	Only for one categorical mediator, set to NULL for numerical mediator or multiple mediator. Reference value of the mediator, where the mediator is evaluated at its refrence value.
stabilized	Stabilized weights, TRUE(default) or FALSE.
R	The number of bootstrap replicates. Default is 1000.

References

Tchetgen Tchetgen, E. J. (2013). Inverse odds ratio‐weighted estimation for causal mediation analysis. Statistics in medicine, 32(26), 4567-4580. DOI:10.1002/sim.5864

Examples

data(lipdat)
dtbase <- lipdat[lipdat$time == 0, ]
out <- iorw(coxph(Surv(os, cvd) ~ bmi + age0 + smoke, data = dtbase),
exposure   = "smoke",
mediator   = "hdl",
family     = "binomial")

summary(out)

adayim/causalMed documentation built on June 2, 2020, 4:11 p.m.