Estimate: Estimate

In podTockom/medltmle: Longitudinal Mediation Analysis with Time-varying Mediators

Estimate

Description

Run GLM or SuperLearner to obtain an estimate for the current node.

Usage

Estimate(
  inputs,
  form,
  subs,
  family,
  type,
  nodes,
  Qstar.kplus1,
  cur.node,
  calc.meanL,
  called.from.estimate.g,
  regimes.meanL,
  regimes.with.positive.weight,
  CSE_I = FALSE,
  CSE_Z = FALSE,
  regimes_add = NULL
)

Arguments

inputs	Output of CreateMedInputs.
form	Q form for current node.
subs	Logical indicating uncensored and non-deterministic samples.
family	a description of the error distribution and link function to be used in the model.
type	"response" or "link"
nodes	List with index for each node subset. Output of CreateNodes.
Qstar.kplus1	Estimate of the expectation with respect to the distribution of the node one ahead of the current node given the past (dimension: n x num.regimes).
cur.node	Node being estimated.
calc.meanL	Estimate conditional density of A using mean of covariates. This is a useful option if there are NAs in the regime, or want to estimate variance.
called.from.estimate.g	Logical TO DO
regimes.meanL	TO DO
regimes.with.positive.weight	Regimes with positve weights. Defaults to 1:num.regimes.
CSE_I	Logical indicating whether this is estimation with instrumental variable.
CSE_Z	Logical indicating whether this is a Z estimation for the data-dependent parameter.
regimes_add	If CSE_Z is TRUE, must specify at what value the node after the mediator should be estimated at.

Value

Returns predicted values for the fit as well as the fit as well as indicators for which samples are deterministic and their values. If specified, it also returns the probability of A=1 given mean L.

podTockom/medltmle documentation built on Aug. 9, 2022, 9:32 a.m.