Minimal CATE demo data (oracle nuisances + pseudo-outcome)

Generates a small synthetic dataset for experimenting with the CATE score. It returns $X$, $D$, $Y$, along with oracle nuisance quantities: $$ m(x) \;=\; \mathbb{E}[Y \mid X=x] \;=\; m_{0}(x) \;+\; e(x)\,\tau(x) \, , \qquad e(x) \;=\; \Pr(D=1 \mid X=x) \, , $$ where $m_{0}(x) = \mathbb{E}[Y \mid X=x, D=0]$ and $\tau(x)$ is the treatment effect. It also returns the R-learner–style pseudo-outcome $$ \rho \;=\; \frac{(D - e)\,(Y - m)}{e\,(1 - e)} \, . $$

Usage

cate_demo_data(
  n = 200,
  p = 2,
  design = c("simple", "hetero", "bump"),
  sigma = 0.5,
  clip = 0.001,
  seed = NULL,
  x0 = NULL,
  inject = NULL,
  inject_col = "rho_oracle",
  idx_inject = 3L
)

Arguments

n: Integer sample size.
p: Integer feature dimension (recommend $p \ge 2$).
design: Character; one of "simple", "hetero", "bump" for $\tau(x)$.
sigma: Noise standard deviation in $Y$.
clip: Propensity clipping in $(0,\,0.5)$; applied to $e(x)$.
seed: Optional integer seed for reproducibility.
x0: Optional numeric length-$p$ target point (defaults to the zero vector).
inject: Optional, one of NULL, "na", "inf"; if provided, inject a single non-finite value into one of the returned columns (see inject_col) at row idx_inject; useful for testing runtime guards.
inject_col: Character; which column to corrupt when inject is not NULL. One of "Y", "rho_oracle", "e_oracle", "m_oracle", "m0_oracle", "tau_oracle". Default "rho_oracle".
idx_inject: Integer index to corrupt (default 3).

Value

A list with:

data: a data.frame with columns X1..Xp, D, Y, and oracle nuisances e_oracle (=$e(X)$), m_oracle (=$m(X)$), m0_oracle (=$m_{0}(X)$), tau_oracle (=$\tau(X)$), rho_oracle (=$\rho$).
x0: the target point used (length $p$).
truth: list of closures tau_fn(x), e_fn(x), m0_fn(x), m_fn(x) to evaluate ground truth at any $x$.
tau_x0: the true $\tau(x_{0})$.

Details

No model fitting is performed; all quantities come from the DGP.

Examples

demo <- cate_demo_data(n = 200, p = 2, design = "hetero", seed = 1)
str(demo$data)
#> 'data.frame':	200 obs. of  9 variables:
#>  $ X1        : num  -0.626 0.184 -0.836 1.595 0.33 ...
#>  $ X2        : num  0.409 1.689 1.587 -0.331 -2.285 ...
#>  $ D         : int  1 0 1 0 1 0 0 0 1 0 ...
#>  $ Y         : num  0.21 -0.654 -1.885 2.932 4.772 ...
#>  $ e_oracle  : num  0.373 0.32 0.23 0.724 0.787 ...
#>  $ m_oracle  : num  0.0414 -0.3159 -1.3926 3.6498 4.1356 ...
#>  $ m0_oracle : num  -0.0359 -0.5052 -1.4222 2.9262 3.6147 ...
#>  $ tau_oracle: num  0.207 0.591 0.129 1 0.662 ...
#>  $ rho_oracle: num  0.451 0.497 -2.144 2.599 0.809 ...
demo$tau_x0
#> [1] 0.5