Controllable Generative Sandbox for Causal Inference

TL;DR

CausalMix is a novel variational generative framework that creates realistic synthetic causal data with explicit control over key causal factors, enabling better method validation and study design in causal inference.

Contribution

It introduces CausalMix, which combines Gaussian mixture priors with data-type-specific decoders and causal controls, balancing realism and controllability in synthetic data generation.

Findings

Achieves state-of-the-art metrics on mixed-type tabular data

Provides stable and fine-grained causal control

Demonstrates utility in safety study and power analysis

Abstract

Method validation and study design in causal inference rely on synthetic data with known counterfactuals. Existing simulators trade off distributional realism, the ability to capture mixed-type and multimodal tabular data, against causal controllability, including explicit control over overlap, unmeasured confounding, and treatment effect heterogeneity. We introduce CausalMix, a variational generative framework that closes this gap by coupling a mixture of Gaussian latent priors with data-type-specific decoders for continuous, binary, and categorical variables. The model incorporates explicit causal controls: an overlap regularizer shaping propensity-score distributions, alongside direct parameterizations of confounding strength and effect heterogeneity. This unified objective preserves fidelity to the observed data while enabling factorial manipulation of causal mechanisms, allowing overlap, confounding strength, and treatment effect heterogeneity to be varied independently at design time. Across benchmarks, CausalMix achieves state-of-the-art distributional metrics on mixed-type tables while providing stable, fine-grained causal control. We demonstrate practical utility in a comparative safety study of metastatic castration-resistant prostate cancer treatments, using CausalMix to compare estimators under calibrated data-generating processes, tune hyperparameters, and conduct simulation-based power analyses under targeted treatment effect heterogeneity scenarios.