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

medltmle

R Documentation

medltmle

Description

Estimates parameters for longitudinal mediation analysis with time-varying mediators.

Usage

medltmle(
  data,
  Anodes,
  Znodes,
  Cnodes = NULL,
  Lnodes = NULL,
  Ynodes,
  Inodes = NULL,
  W2nodes = NULL,
  Dnodes = NULL,
  survivalOutcome = NULL,
  QLform = NULL,
  QZform = NULL,
  gform = NULL,
  qzform = NULL,
  qLform = NULL,
  abar,
  abar.prime,
  rule = NULL,
  gbounds = c(0.01, 1),
  Yrange = NULL,
  deterministic.g.function = NULL,
  deterministic.Q.function = NULL,
  stratify = FALSE,
  SL.library = NULL,
  estimate.time = TRUE,
  gcomp = FALSE,
  iptw.only = FALSE,
  IC.variance.only = FALSE,
  observation.weights = NULL,
  CSE,
  time.end,
  past = 1,
  YisL = TRUE
)

Arguments

`data`	Dataframe containing the data in a wide format.
`Anodes`	names of columns containing A covariates (exposure) (character).
`Znodes`	names of columns containing Z covariates (mediator) (character).
`Cnodes`	names of columns containing C covariates (censoring) (character).
`Lnodes`	names of columns containing L covariates (covariate) (character).
`Ynodes`	names of columns containing Y covariates (outcome) (character).
`Inodes`	names of columns containing I covariates (instrument) (character).
`W2nodes`	names of columns containing W2 covariates (baseline covariates in need of fluctuation) (character).
`Dnodes`	names of columns containing D covariates (death indicator) (character).
`survivalOutcome`	If TRUE, then Y nodes are indicators of an event, and if Y at some time point is 1, then all following should be 1. Required to be TRUE or FALSE if outcomes are binary and there are multiple Ynodes.
`QLform`	character vector of regression formulas for Q corresponding to L covariates.
`QZform`	character vector of regression formulas for Q corresponding to Z covariates.
`gform`	character vector of regression formulas for g or a matrix/array of prob(A=1).
`qzform`	g form for Z covariates.
`qLform`	g form for L covariates.
`abar`	binary vector (numAnodes x 1) or matrix (n x numAnodes) of counterfactual treatment or a list of length 2.
`abar.prime`	binary vector (numAnodes x 1) or matrix (n x numAnodes) of counterfactual treatment or a list of length 2.
`rule`	a function to be applied to each row (a named vector) of data that returns a numeric vector of length numAnodes.
`gbounds`	lower and upper bounds on estimated cumulative probabilities for g-factors. Vector of length 2, order unimportant.
`Yrange`	NULL or a numerical vector where the min and max of Yrange specify the range of all Y nodes.
`deterministic.g.function`	optional information on A and C nodes that are given deterministically. Default NULL indicates no deterministic links.
`deterministic.Q.function`	optional information on Q given deterministically. See 'Details'. Default NULL indicates no deterministic links.
`stratify`	if TRUE stratify on following abar when estimating Q and g. If FALSE, pool over abar.
`SL.library`	optional character vector of libraries to pass to SuperLearner. NULL indicates glm should be called instead of SuperLearner. 'default' indicates a standard set of libraries. May be separately specified for Q and g.
`estimate.time`	if TRUE, run an initial estimate using only 50 observations and use this to print a very rough estimate of the total time to completion. No action if there are fewer than 50 observations.
`gcomp`	if TRUE, run the maximum likelihood based G-computation estimate instead of TMLE.
`iptw.only`	by default (iptw.only = FALSE), both TMLE and IPTW are run in ltmle and ltmleMSM. If iptw.only = TRUE, only IPTW is run, which is faster.
`IC.variance.only`	Only estimate variance through the influence curve
`observation.weights`	observation (sampling) weights. Vector of length n. If NULL, assumed to be all 1.
`CSE`	Logical specifying if the estimand is estimated by fully conditioning on the past (TRUE), or with the data-dependent estimate (FALSE).
`time.end`	How many time points in the longitudinal data?
`past`	Number indicating Markov order for the conditional densities.
`YisL`	Logical indicating whether Y is a function of time-varying covariate.