PYCD-LINGAM: A PYTHON FRAMEWORK FOR CAUSAL INFERENCE WITH NON-GAUSSIAN LINEAR MODELS

PyCD-LiNGAM is an advanced Python framework designed to facilitate causal inference in observational data using non-Gaussian linear models. Building upon the foundational principles of the Linear Non-Gaussian Acyclic Model (LiNGAM), this framework offers a robust suite of tools for uncovering causal structures in datasets where conventional Gaussian assumptions fail to capture latent dependencies. PyCD-LiNGAM provides efficient implementations of DirectLiNGAM, ICA-LiNGAM, and adaptive algorithms that exploit higher-order statistical properties to reliably identify causal ordering and estimate connection strengths among variables. The framework integrates seamlessly with popular scientific computing libraries, enabling practitioners to perform end-to-end causal discovery, visualize directed acyclic graphs, and assess model fit through rigorous statistical criteria. Benchmark experiments demonstrate that PyCD-LiNGAM achieves high accuracy and scalability across synthetic and real-world datasets, outperforming baseline methods in identifying true causal relationships under non-Gaussian noise. By lowering the barrier to applying state-of-the-art causal inference techniques, PyCD-LiNGAM empowers researchers and data scientists in fields such as econometrics, neuroscience, genomics, and social sciences to derive actionable insights about underlying causal mechanisms.