Direct effect with the mediator drawn from a fixed reference arm

Fixing the mediator distribution at the reference arm rather than at each person own counterfactual value is what makes this identified when an exposure-induced confounder sits between mediator and outcome. The mediator model is fitted on the reference arm and imposed on all.

Usage

Tmlnde2(y, D, M, X, a_ref = 0)

Arguments

y

Outcome.

D

Binary treatment.

M

Mediator.

X

Baseline covariates.

a_ref

Arm whose mediator distribution is imposed.

Value

List with estimate, se, eps, m_shift, n.

Details

Formula: E[Y(1, M_a*) - Y(0, M_a*)], targeted with H = D/g - (1 - D)/(1 - g).

References

Vansteelandt, S. & Daniel, R. M. (2017). Epidemiology 28:258-265.

Examples

Tmlnde2(y = c(1, 2, 3, 4, 5, 6, 7, 8), D = 5L, M = 5L, X = c(1, 2, 3, 4, 5, 6, 7, 8))
#> $estimate
#> [1] 4.563542e+30
#> 
#> $se
#> [1] 0
#> 
#> $eps
#> [1] 1.090934e+14
#> 
#> $m_shift
#> [1] -5
#> 
#> $n
#> [1] 8
#> 
#> $method
#> [1] "TMLE for an interventional direct effect"
#> 
#> attr(,"class")
#> [1] "morie_rich_result" "list"