Tutorials

This page gives minimal templates for the three workflows.

Tutorial A: Observational Heterogeneous Effects

library(heteff)
set.seed(1)
n <- 800
x1 <- rnorm(n)
x2 <- rnorm(n)
w <- rbinom(n, 1, plogis(0.5 * x1 - 0.2 * x2))
y <- 0.5 * w + 0.8 * w * (x1 > 0) + x1 + 0.3 * x2 + rnorm(n)

df <- data.frame(
  outcome = y,
  treatment = w,
  x1 = x1,
  x2 = x2
)

fit_obs <- fit_observational_forest(
  data = df,
  outcome = "outcome",
  treatment = "treatment",
  covariates = c("x1", "x2")
)

plot(fit_obs)
head(rank_effects(fit_obs, n = 15))

Tutorial B: Survival Heterogeneous Effects

set.seed(2)
n <- 700
x1 <- rnorm(n)
x2 <- rnorm(n)
w <- rbinom(n, 1, 0.5)
time <- rexp(n, rate = exp(-0.3 * w - 0.2 * x1))
status <- rbinom(n, 1, 0.85)

df_surv <- data.frame(
  time = time,
  status = status,
  treatment = w,
  x1 = x1,
  x2 = x2
)

fit_surv <- fit_survival_forest(
  data = df_surv,
  time = "time",
  status = "status",
  treatment = "treatment",
  covariates = c("x1", "x2"),
  target = "RMST",
  horizon = 5
)

plot(fit_surv)
head(as.data.frame(fit_surv))

Tutorial C: Instrumental Heterogeneous Effects

set.seed(3)
n <- 1000
x1 <- rnorm(n)
x2 <- rnorm(n)
z <- rbinom(n, 1, plogis(0.3 * x1))
w <- 0.7 * z + 0.4 * x1 + rnorm(n)
y <- 1.4 * w + 0.5 * x2 + 0.6 * (x1 > 0) * w + rnorm(n)

df_iv <- data.frame(
  outcome = y,
  treatment = w,
  instrument = z,
  x1 = x1,
  x2 = x2
)

fit_iv <- fit_instrumental_forest(
  data = df_iv,
  outcome = "outcome",
  treatment = "treatment",
  instrument = "instrument",
  covariates = c("x1", "x2")
)

plot(fit_iv)
head(rank_effects(fit_iv, n = 20))

Export Pattern

obs_table <- as.data.frame(fit_obs)
iv_top <- rank_effects(fit_iv, n = 30)

write.csv(obs_table, "observational_effects.csv", row.names = FALSE)
write.csv(iv_top, "instrumental_top30.csv", row.names = FALSE)