setup_plot: Set up information for a `GenericML()` plot

Description

Extract the relevant information for visualizing the point and interval estimates of the generic targets of interest. The generic targets of interest can be (subsets of) the parameters of the BLP, GATES, or CLAN analysis.

Usage

setup_plot(
  x,
  type = "GATES",
  learner = "best",
  CLAN_variable = NULL,
  groups = "all"
)

Arguments

An object of the class "GenericML", as returned by the function GenericML().

type

The analysis whose parameters shall be plotted. Either "GATES", "BLP", or "CLAN". Default is "GATES".

learner

The learner whose results are to be returned. Default is "best" for the best learner as measured by the \(\Lambda\) parameters.

CLAN_variable

Name of the CLAN variable to be plotted. Only applicable if type = "CLAN".

groups

Character vector indicating the per-group parameter estimates that shall be plotted in GATES and CLAN analyses. Default is "all" for all parameters. If there are \(K\) groups, this variable is a subset of c("G1", "G2",...,"GK", "G1-G2", "G1-G2",..., "G1-GK", "GK-G1", "GK-G2",...), where Gk denotes the k-th group. Note that this set depends on the choices of the arguments "diff_GATES" and "diff_CLAN" of the "GenericML" object.

Value

An object of class "setup_plot", which is a list with the following elements.

data_plot: A data frame containing point and interval estimates of the generic target specified in the argument type.
data_BLP: A data frame containing point and interval estimates of the BLP analysis.
confidence_level: The confidence level of the confidence intervals. The confidence level is equal to 1 - 2 * significance_level, which is the adjustment proposed in the paper.

Details

This function is used internally by plot.GenericML(). It may also be useful for users who want to produce a similar plot, but who want more control over what information to display or how to display that information.

Examples

Run this code

# NOT RUN {
if(require("ranger") && require("ggplot2")) {

## generate data
set.seed(1)
n  <- 150                                  # number of observations
p  <- 5                                    # number of covariates
D  <- rbinom(n, 1, 0.5)                    # random treatment assignment
Z  <- matrix(runif(n*p), n, p)             # design matrix
Y0 <- as.numeric(Z %*% rexp(p) + rnorm(n)) # potential outcome without treatment
Y1 <- 2 + Y0                               # potential outcome under treatment
Y  <- ifelse(D == 1, Y1, Y0)               # observed outcome

## name the columns of Z
colnames(Z) <- paste0("V", 1:p)

## specify learners
learners <- c("random_forest")

## perform generic ML inference
# small number of splits to keep computation time low
x <- GenericML(Z, D, Y, learners,
               num_splits = 2,
               parallel = FALSE)

## the plot we wish to replicate
plot(x = x, type = "GATES")

## get the data to plot the GATES estimates
data <- setup_plot(x = x, type = "GATES")

## define variables to appease the R CMD check
group <- estimate <- ci_lower <- ci_upper <- NULL

## replicate the plot(x, type = "GATES")
# for simplicity, we skip aligning the colors
ggplot(mapping = aes(x = group,
                     y = estimate), data = data$data_plot) +
  geom_hline(aes(yintercept = 0),
             color = "black", linetype = "dotted") +
  geom_hline(aes(yintercept = data$data_BLP["beta.1", "estimate"],
                 color = "ATE"),
             linetype = "dashed") +
  geom_hline(aes(yintercept = data$data_BLP["beta.1", "ci_lower"],
                 color = paste0(100*data$confidence_level, "% CI (ATE)")),
             linetype = "dashed")  +
  geom_hline(yintercept = data$data_BLP["beta.1", "ci_upper"],
             linetype = "dashed", color = "red") +
  geom_point(aes(color = paste0("GATES with ",  100*data$confidence_level, "% CI")), size = 3) +
  geom_errorbar(mapping = aes(ymin = ci_lower,
                              ymax = ci_upper))
}

# }