Skip to contents

Draw one sample from a named IV-GMM simulation DGP

Source: R/sim-iv-data.R
sim_iv_data.Rd

Generates a synthetic local IV-GMM dataset with oracle nuisance quantities for the doubly-robust IV moment $$ (Z_k - \pi_k(X))\{Y - \mu_Y(X) - (D - \mu_D(X))\theta\}. $$ The returned data are estimator-free and are intended for simulation benchmarks, examples, and future gmm_dnn() integration.

Usage

sim_iv_data(
  design,
  n,
  kx = 2L,
  q = 2L,
  sigma = 0.5,
  clip = 0.001,
  seed = NULL,
  x0 = NULL,
  ...
)

Arguments

design

Character DGP name. One of "dgp1_iv_justid_linear", "dgp2_iv_homog_overid", "dgp3_iv_heterog_overid", "dgp4_iv_weak_local", "dgp5_iv_correlated_instruments", or "dgp6_iv_nonlinear_firststage".

n

Integer sample size.

kx

Integer covariate dimension, at least 2.

q

Integer number of instruments. Phase-A simulation DGPs support 1 <= q <= 3. The just-identified design always returns q = 1.

sigma

Noise standard deviation in Y.

clip

Clipping bound in (0, 0.5) for binary instrument probabilities.

seed

Optional integer seed for reproducibility.

x0

Optional numeric length-kx evaluation point. When omitted, a design-specific interior point is used.

...

Extra arguments forwarded to the underlying DGP. Currently rho is used by "dgp5_iv_correlated_instruments".

Value

A list with:

  • data: a data.frame with X1..Xkx, Y, D, Z1..Zq, oracle nuisance columns piZ*_oracle, muY_oracle, muD_oracle, and the target column theta_oracle.

  • x0: numeric length-kx evaluation point.

  • q: realized number of instruments.

  • theta_x0: true local IV target at x0.

  • truth: named list of oracle functions.

Examples

dgp <- sim_iv_data("dgp2_iv_homog_overid", n = 200, kx = 2, q = 2, seed = 1)
head(dgp$data)
#>           X1         X2          Y D Z1 Z2 piZ1_oracle piZ2_oracle muY_oracle
#> 1 -0.6264538  0.4094018 -0.4726267 0  1  0   0.3842354   0.4325121  0.1585522
#> 2  0.1836433  1.6888733  2.4378363 1  0  1   0.3756758   0.4236169  0.5734972
#> 3 -0.8356286  1.5865884 -0.3643560 1  1  0   0.3017208   0.3454461 -0.6673652
#> 4  1.5952808 -0.3309078  2.5523677 0  0  0   0.6203597   0.6662061  3.5175028
#> 5  0.3295078 -2.2852355  2.8350715 1  1  1   0.6310297   0.6762599  2.9869289
#> 6 -0.8204684  2.4976616 -1.8084327 0  0  0   0.2570078   0.2970095 -1.0987363
#>   muD_oracle theta_oracle
#> 1  0.3538041   -0.0290925
#> 2  0.3493445    0.6706575
#> 3  0.2588385   -0.1485186
#> 4  0.6392690    1.1838023
#> 5  0.6433009    0.8002528
#> 6  0.2083843   -0.1412635
dgp$theta_x0
#> [1] 1.211738