Synthia

Spec sheet

Purpose

Synthia is an open source Python package for modeling univariate and multivariate data, parameterizing data, and manipulating marginal distributions.Read 4 Oct 2026 · dmey.github.io

Intended users

It is designed for scientists and practitioners working with labelled multivariate data typical of computational sciences.Read 4 Oct 2026 · dmey.github.io

License

Synthia is licensed under the MIT License, which permits use, copying, modification, distribution, sublicensing, and sale subject to its stated conditions.Read 4 Oct 2026 · dmey.github.io

Data integration

Synthia's API natively handles xarray labelled arrays and datasets.Read 4 Oct 2026 · dmey.github.io

Distribution

The package can be installed with pip install synthia, or with pip install synthia[full] to include the optional dependency.Read 4 Oct 2026 · dmey.github.io

Example use

The quickstart demonstrates generating synthetic atmospheric temperature profiles using Gaussian copula and fPCA models.Read 4 Oct 2026 · dmey.github.io

Generation methods

It supports fPCA, Gaussian copula, and vine copula methods for multivariate data generation.Read 4 Oct 2026 · dmey.github.io

Implementation

The fPCA and Gaussian copula implementations are in Python, while vine computations use pyvinecopulib bindings to the C++ vinecopulib library.Read 4 Oct 2026 · dmey.github.io

Optional dependency

pyvinecopulib is an optional dependency for vine copula and quasirandom number support.Read 4 Oct 2026 · dmey.github.io

Requirements

Installation requires Python 3.8 or later, NumPy, SciPy, and xarray.Read 4 Oct 2026 · dmey.github.io

Variable types

The documentation says its methods support continuous variables and vine copulas also support discrete and categorical variables.Read 4 Oct 2026 · dmey.github.io

Warranty

The MIT license states that the software is provided as is and disclaims warranties and liability as specified in the license.Read 4 Oct 2026 · dmey.github.io

Where Specifiction places it

• Synthetic Data Generation Software#28 of 296.0