Lalonde

whynot.simulators.lalonde.experiments.run_lalonde(num_samples, hidden_dim, alpha_scale, seed=None, parallelize=True, show_progress=False)[source]

Generate data from the LaLonde dataset with a random response function.

The covariates and treatment are both specified by the dataset, and the response function is a random 2-layer neural network with ReLu.

Parameters:
  • num_samples (int) – This parameter is ignored since the LaLonde dataset size is fixed.
  • hidden_dim (int) – Hidden dimension of the relu network.
  • alpha_scale (float) – Standard deviation of the final layer weights.
  • seed (int) – Random seed used for all internal randomness
  • parallelize (bool) – Ignored, but included for consistency with GenericExperiment API.
  • show_progress (False) – Ignored, but included for consistency with GenericExperiment API.

Experiments

Basic set of experiments on the LaLonde dataset.

whynot.simulators.lalonde.experiments.RandomResponse = <whynot.framework.GenericExperiment object>

Experiment simulating an outcome function on top of fixed LaLonde covariates.