dagsampler documentationΒΆ
Minimal Python package for generating synthetic data from causal DAGs.
Contents
- Overview
- Model Formulations
- Notation
- Graph Model
- Node Types
- Exogenous Nodes (\(\mathrm{Pa}(j)=\varnothing\))
- Endogenous Continuous Nodes
- Random structural weights
- Post-nonlinear transform
- Endogenous Binary Nodes
- Endogenous Categorical Nodes
- Compatibility Matrix
- Forced uniform marginals
- Random node-type assignment
- Categorical parents in metric forms
- Stratum means reproducibility
- CI Oracle (Ground Truth)
- Configuration Examples
- Seeding
- Minimal custom DAG
- Random DAG
- Random weights with near-zero exclusion (signal-strength control)
- Categorical parent with metric form policy override
- Exogenous node distributions
- Continuous child with linear / polynomial / interaction
- Continuous child with sigmoid / cos / sin
- Post-nonlinear transform
- Noise model variants
- Heavy-tailed and uniform additive noise
- Forced uniform marginals
- Binary child
- Categorical child (logistic softmax)
- Continuous to categorical (threshold)
- Categorical to continuous (stratum-specific means)
- Mixed parents under stratum_means
- CI oracle output
- simulate() return value
- Template Configurations
- Usage
- API Reference