Benchmark estimators across pleiotropy and confounding scenarios

For each cell in the grid (pleiotropy strength x confounding strength), generates n_iter synthetic datasets with run_single_iteration(), runs every estimator, and summarises bias / RMSE / power. Two arms are run per cell:

Usage

gan_pleiotropy_sensitivity(
  trained_gan = NULL,
  pleio_grid = c(0, 0.05, 0.1),
  conf_grid = c(0.2, 0.5, 0.8),
  tau = 0.25,
  nc_model = "proxy",
  n_iter = 50,
  n_samples = 500,
  n_features = 10,
  coverage = 0.7,
  k = 1,
  base_seed = 900,
  n_cores = 1
)

Arguments

trained_gan

An iconic_gan (or NULL to use default texture).

pleio_grid

Horizontal-pleiotropy strengths (direct G -> Y coefficients) to sweep. Default c(0, 0.05, 0.10).

conf_grid

Confounding-strength values. Default c(0.2, 0.5, 0.8).

tau

True total effect for the alternative arm. Default 0.25.

nc_model

Negative-control model (function or name). Default "proxy".

n_iter

Replicates per cell per arm. Default 50.

n_samples

Samples per replicate. Default 500.

n_features

Features per replicate. Default 10.

coverage

Negative-control coverage. Default 0.7.

k

Number of latent confounders. Default 1.

base_seed

Base RNG seed. Default 900.

n_cores

Parallel workers across replicates. Default 1.

Value

A list with summary (one row per cell x arm x method, with pleio, conf_strength, arm, true_total, and the columns from summarise_results()) and grid.

Details

• alternative (effect_size = tau): the true total effect is tau, so power is the empirical power to detect it.

• null (effect_size = 0): the true total effect is 0, so power is the empirical Type I error rate.

The pleio parameter adds a direct G -> Y path of the requested strength, violating the exclusion restriction. IV/2SLS is consistent only when pleio = 0; any pleio > 0 introduces bias that does not shrink with sample size.

Examples

sens <- gan_pleiotropy_sensitivity(NULL,
  pleio_grid = c(0, 0.10), conf_grid = 0.8,
  n_iter = 2, n_samples = 100, n_features = 5)
head(sens$summary)
#>   pleio conf_strength  arm true_total method       mean     median          sd
#> 1     0           0.8  alt       0.25  UNADJ  0.5600946  0.5828345 0.072883744
#> 2     0           0.8  alt       0.25 DIRECT  0.3634369  0.3843584 0.041171431
#> 3     0           0.8  alt       0.25   COCA -0.2458452 -0.1827060 0.146524016
#> 4     0           0.8  alt       0.25 IV2SLS  0.1550511  0.1586644 0.009041299
#> 5     0           0.8  alt       0.25    PGC  0.2625661  0.2631845 0.017282523
#> 6     0           0.8 null       0.00  UNADJ  0.2969582  0.2921351 0.047463252
#>          bias   abs_bias      rmse power  n
#> 1  0.31009459 0.31009459 0.3177098   1.0 10
#> 2  0.11343695 0.11343695 0.1199730   1.0 10
#> 3 -0.49584516 0.49584516 0.5149610   0.4 10
#> 4 -0.09494887 0.09494887 0.0953355   0.2 10
#> 5  0.01256611 0.01256611 0.0206573   1.0 10
#> 6  0.29695818 0.29695818 0.3003525   1.0 10