sample_design

obsidian.experiment.advanced_design.sample_design(seed, n_samples, continuous_params, conditional_subparameters, subparam_mapping=None, optimize_categories=False, n_category_trials=100, corr_threshold=0.01)

Generate a complete experimental design by LHS-sampling all continuous and categorical parameters.

Parameters:

• seed – Random seed for reproducibility.

• n_samples – Number of rows to generate.

• continuous_params – Continuous parameter specifications (see _get_param_levels).

• conditional_subparameters – Conditional categorical/subparameter specifications.

• subparam_mapping – Dict mapping categorical variables to their subparameters. Inferred automatically if not provided.

• optimize_categories – Whether to optimize the category assignment for the variable that has a subparam mapping to reduce inter-category correlation. Defaults to False.

• n_category_trials – Number of random assignments evaluated during category optimization. Defaults to 100.

• corr_threshold – Early-exit correlation threshold for category optimization. Defaults to 0.01.

Returns:

The generated design with appropriately rounded values.

Return type:

pd.DataFrame

Note

When optimize_categories=True, only the first subparam-mapped category (as determined by subparam_mapping) is optimized. Additional categorical variables are assigned with a single random draw.