lmtp_contrast function

Perform Contrasts of LMTP Fits

Estimates contrasts of multiple LMTP fits compared to either a known reference value or a reference LMTP fit.

lmtp_contrast(..., ref, type = c("additive", "rr", "or"))

Arguments

• ...: One or more objects of class lmtp.
• ref: A reference value or another lmtp fit to compare all other fits against.
• type: The contrasts of interest. Options are "additive" (the default), "rr", and "or".

Returns

A list of class lmtp_contrast containing the following components:

• type: The type of contrast performed.

• null: The null hypothesis.

• vals: A dataframe containing the contrasts estimates, standard errors, and confidence intervals.

• eifs: Un-centered estimated influence functions for contrasts estimated.

Examples

a <- c("A1", "A2")
  nodes <- list(c("L1"), c("L2"))
  cens <- c("C1", "C2")
  y <- "Y"

  # mean population outcome
  psi_null <- lmtp_tmle(sim_cens, a, y, time_vary = nodes, cens = cens, shift = NULL, folds = 1)

  # treatment rule, everyone is increased by 0.5
  d <- function(data, x) data[[x]] + 0.5
  psi_rule1 <- lmtp_tmle(sim_cens, a, y, time_vary = nodes, cens = cens,
                         shift = d, folds = 1, mtp = TRUE)

  # treatment rule, everyone is decreased by 0.5
  d <- function(data, x) data[[x]] - 0.5
  psi_rule2 <- lmtp_tmle(sim_cens, a, y, time_vary = nodes, cens = cens,
                         shift = d, folds = 1, mtp = TRUE)

  # Example 1.1
  # Additive effect of rule 1 compared to a known constant
  lmtp_contrast(psi_rule1, ref = 0.9)

  # Example 1.2
  # Additive effect of rule 1 compared to the population mean outcome
  lmtp_contrast(psi_rule1, ref = psi_null)

  # Example 1.3
  # Additive effects of rule 1 and 2 compared to the population mean outcome
  lmtp_contrast(psi_rule1, psi_rule2, ref = psi_null)

  # Example 1.4
  # Relative risk of rule 1 compared to observed exposure
  lmtp_contrast(psi_rule1, ref = psi_null, type = "rr")

  # Example 1.5
  # Odds of rule 1 compared to observed exposure
  lmtp_contrast(psi_rule1, ref = psi_null, type = "or")