Function to evaluate the canonical gradient of the direct effect at a given set of estimated nuisance parameters and for each observation.

Function to evaluate the canonical gradient of the direct effect at a given set of estimated nuisance parameters and for each observation.

evaluate_eif_direct(
  Y,
  A,
  M1,
  M2,
  Qbar,
  Q_M,
  Qbarbar,
  gn,
  a,
  a_star,
  use_conditional,
  all_mediator_values,
  ...
)

Arguments

Y	The outcome
A	The treatment
Qbar	Outcome regression list
Q_M	Mediator distribution list
Qbarbar	A list; needs to have entries named M1_M2_a and M1_star_M2_star_a_star corresponding to, respectively, the outcome regression under A = a, marginalized with respect to joint mediator distribution given C and A = a, and the outcome regression under A = a_star, marginalized with respect to the joint mediator distribution given C and A = a_star.
gn	List of propensity scores
a	Treatment value
a_star	Referent treatment value
use_conditional	Boolean. Should the conditional_total_effect be used (generally used for when evaluating EIF of targeted nuisance parameters, since there we target directly the difference)