Total effect by TMLE, with the mediator held out of the model

The natural total effect is the direct and indirect effects added back together, so the mediator must not enter the outcome model: conditioning on it would block the path being counted. The mediator is used only for the contrast report.

Usage

Tmlnte(y, D, M, X)

Arguments

y

Outcome.

D

Binary treatment.

M

Mediator; reported on, not adjusted for.

X

Baseline covariates.

Value

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

Details

Formula: NTE = E[Y(1) - Y(0)] = NDE + NIE, targeted with H = D/g - (1 - D)/(1 - g).

References

VanderWeele (2015), Explanation in Causal Inference, ch 2; van der Laan & Rubin (2006) IJB 2(1):11.

Examples

Tmlnte(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] -8.081304e+15
#> 
#> $se
#> [1] 142.9779
#> 
#> $eps
#> [1] -9.818407e+13
#> 
#> $nde_naive
#> [1] -9.098739e+30
#> 
#> $n
#> [1] 8
#> 
#> $method
#> [1] "TMLE for the natural total effect"
#> 
#> attr(,"class")
#> [1] "morie_rich_result" "list"