plot_uplift: Creates an uplift plot of cumulative differential...

In uplifteval: Uplift Model Evaluation with Plots and Metrics

Description

Creates an uplift plot of cumulative differential treatment/control outcomes versus model score. Also provides a selection of metrics: max uplift as pct of total control outcome, optimum users targeted and optimum score targeting range.

Usage

plot_uplift(p1, W, Y, ns = min(table(W)), n_bs = 1, W_label = W,
  p0 = rep(0, length(p1)), balanced = TRUE, replace = TRUE,
  x_interval = 0.1, ...)

Arguments

p1	numeric vector of uplift predictions; can also be predicted outcomes for treated case (in this case p0 should contain predicted outcomes for the control case)
W	binary vector 1,0 of treatment assignments
Y	numeric vector of responses
ns	integer number of samples per bootstrap iteration; default min(table(W))
n_bs	integer number of bootstrap iterations
W_label	optional labels for the treatment options (default W)
p0	optional numeric vector of predicted outcomes for control case
balanced	optional boolean whether to sample equal proportions from treatment and control cases; default TRUE
replace	optional boolean whether to use replacement when sampling; default TRUE
x_interval	optional numeric the interval with which to split the
...	additional arguments (unused) x-axis

Examples

set.seed(0)
rl <- function(x){
  round(1/(1+exp(-x)))
}
n <- 2000; p <- 3
beta <- -0.5
X <- matrix(rnorm(n*p), n, p)
W <- rbinom(n, 1, 0.5)
Y <- rl(pmax(beta+X[,1], 0) * W + X[,2])
p1 <- 1/(1+exp(-(beta+X[,1])))
plot_uplift(p1, W, Y, n_bs=20, x_interval = 0.05, balanced = TRUE)


set.seed(0)
n <- 2000; p <- 3
beta <- -0.5
X <- matrix(rnorm(n*p), n, p)
W <- rbinom(n, 1, 0.5)
Y <- pmax(beta+X[,1], 0) * W + X[,2]
p1 <- 1/(1+exp(-(beta+X[,1])))
plot_uplift(p1, W, Y, n_bs=20, x_interval = 0.05, balanced = TRUE)



library(grf)
set.seed(123)

rl <- function(x){
  round(1/(1+exp(-x)))
}
n = 2000; p = 10
X = matrix(rnorm(n*p), n, p)
W = rbinom(n, 1, 0.2)
Y = rl(rl(X[,1]) * W - rl(X[,3]) * W + rnorm(n))
tau.forest = causal_forest(X, Y, W)
tau.hat = predict(tau.forest, X)
plot_uplift(tau.hat$predictions, W, Y, n_bs=20, x_interval = 0.05, balanced = FALSE)
plot_uplift(tau.hat$predictions, W, Y, n_bs=20, x_interval = 0.05, balanced = TRUE)

uplifteval documentation built on June 15, 2019, 9:03 a.m.