🔭 astrophysics

CosmoFit: A Graphical User Interface for Bayesian Cosmological Parameter Estimation with User-Defined H(z) Models

This paper introduces CosmoFit, a transparent and reproducible desktop graphical interface that simplifies Bayesian cosmological parameter estimation by integrating Cobaya and GetDist to allow users to define custom expansion histories, select observational likelihoods, and perform MCMC analyses without manual programming, while demonstrating its accuracy through rigorous testing and a wCDM analysis example.

Original authors: Giovany Cruz

Published 2026-08-26

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Giovany Cruz

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The universe is not static; it is expanding, stretching the space between galaxies as time moves forward. To understand how fast this expansion is happening and how it has changed over billions of years, astronomers rely on a concept called the expansion history. This is essentially a record of how the rate of cosmic growth has evolved from the early universe to the present day. Scientists test different theories about what drives this expansion—whether it is a constant force, a changing energy field, or something else entirely—by comparing their mathematical predictions against real observations. These observations come from powerful tools like Type Ia supernovae, which act as cosmic mile markers, and cosmic chronometers, which measure the age of galaxies to determine the expansion rate directly. However, turning these observations into a clear answer about the nature of the universe usually requires a complex, manual process of writing code, configuring software, and running massive statistical calculations. This technical barrier often keeps students and researchers from quickly testing their own ideas about how the universe works.

A new tool called CosmoFit aims to remove that barrier without sacrificing scientific rigor. Developed by researcher Giovany Cruz, this desktop application acts as a graphical guide for a process known as Bayesian parameter estimation. In simple terms, this is a method of using data to figure out which values for the universe's properties are most likely to be true. Instead of forcing a user to write lines of code to tell the computer what to do, CosmoFit provides a visual interface where a researcher can type in a specific formula for the expansion history, select the data they want to use, and press a button to start the analysis. The software then handles the heavy lifting behind the scenes, running thousands of simulations to find the best fit between the user's theory and the actual observations.

The core innovation of CosmoFit is how it balances ease of use with transparency. Many tools that simplify scientific work do so by hiding the details, which can make it difficult to verify the results or understand exactly what assumptions were made. CosmoFit avoids this "black box" approach. When a user defines a model for the expansion rate, the software checks the math for errors before starting. It ensures the formula makes sense and does not produce impossible numbers. Once the user is ready, the application automatically translates their visual choices into the precise configuration files required by a powerful, established engine called Cobaya. This engine performs the actual statistical sampling, a process that involves generating random chains of data to map out the probability of different outcomes. Crucially, CosmoFit saves every single file created during this process, from the original user input to the final data logs. This means that anyone can look at the results, repeat the calculation, or inspect the exact steps taken, ensuring that the science remains open and reproducible.

To prove that this new interface works correctly, the author subjected it to a series of strict tests. First, the software was tested with synthetic data, where the true answer was already known. The program successfully recovered the correct values for the universe's expansion rate and matter density, confirming that the workflow functions as intended. Next, the results generated by CosmoFit were compared directly against results from a standard, manual analysis performed without the graphical interface. The numbers matched almost perfectly, with differences in the final answers being less than one percent. This high level of agreement demonstrates that the graphical tool does not alter the underlying science; it simply provides a more accessible way to reach the same conclusions.

The paper also includes a practical demonstration using real data from the Pantheon+SH0ES supernova compilation. In this example, a researcher used the interface to test a specific model of the universe where the expansion is driven by a mysterious energy field with a constant strength. By entering the formula for this model and selecting the supernova data, the software ran the analysis and produced a detailed report. The results included a clear visualization of the most likely values for the expansion rate and the density of matter, along with a chart showing how these two factors relate to each other. The analysis showed that the data is consistent with a standard model of the universe, but the primary achievement here was not the specific result itself. Rather, it was the demonstration that a researcher could move from defining a theory to seeing a professional-grade statistical result entirely within a single, user-friendly environment.

CosmoFit is currently designed to study the background expansion of the universe, meaning it looks at the overall growth of space rather than the complex clumping of matter or the behavior of light in the early universe. It supports specific types of data, including measurements from cosmic chronometers and several major supernova datasets. While it does not yet handle every possible type of cosmological observation, it covers the most common methods used to study the expansion history. The software is open-source, meaning the code is available for anyone to inspect, modify, or build upon. By lowering the technical threshold for entry, CosmoFit allows students and researchers to focus more on the physics of the universe and less on the mechanics of the software, making the process of testing new cosmological ideas faster and more accessible to a wider community.

Technical Summary of CosmoFit: A Graphical User Interface for Bayesian Cosmological Parameter Estimation

Problem Statement
Modern cosmological inference relies heavily on Bayesian parameter estimation using Markov Chain Monte Carlo (MCMC) methods. While mature computational frameworks such as CosmoMC, MontePython, and Cobaya provide flexible environments for defining models and likelihoods, their direct use typically requires significant familiarity with programming, command-line tools, configuration file syntax, and package management. This technical barrier can hinder students, educators, and researchers who wish to test specific background cosmological models—particularly those defined by a modified expansion history $H(z)$ —without the overhead of building a complete computational pipeline. Furthermore, while graphical interfaces exist, they must balance usability with the rigorous requirements of scientific transparency, reproducibility, and the avoidance of unrestricted code execution.

Methodology
The author presents CosmoFit, an open-source desktop graphical interface designed to streamline Bayesian parameter inference for homogeneous, isotropic, and spatially flat background cosmology. The application acts as an orchestration layer that translates user inputs into valid configurations for Cobaya (for MCMC sampling) and GetDist (for posterior analysis and visualization).

Key methodological components include:

Restricted Mathematical Parser: Users define the expansion history $H(z)$ via a mathematical expression. The system employs a restricted parser that validates syntax, checks for undefined domains (e.g., division by zero, complex values), and evaluates the model numerically over a redshift interval before sampling begins. This prevents arbitrary code execution while allowing flexible model definitions.
Workflow Automation: The interface manages the entire workflow: defining fixed and sampled parameters, assigning prior intervals and proposal scales, selecting observational likelihoods, and launching MCMC runs.
Isolated Execution: MCMC calculations are executed in a separate worker process. This prevents long-running numerical tasks from blocking the graphical interface and allows for the cancellation of runs or inspection of logs without closing the application.
Reproducibility: Every run generates a dedicated output directory containing the original user input, a normalized internal configuration, the generated Cobaya input file, metadata, execution logs, MCMC chains, and numerical summaries. This ensures that any analysis can be reconstructed or inspected independently of the GUI.
Supported Likelihoods: The current version supports direct measurements of $H(z)$ from cosmic chronometers and official Cobaya likelihoods for the Pantheon+, Pantheon+SH0ES, and Union3 Type Ia supernova compilations. It does not currently support cosmological perturbations or Boltzmann codes (e.g., CLASS, CAMB).

Key Contributions
The paper identifies seven principal contributions:

A desktop graphical workflow for configuring Bayesian analyses of user-defined cosmological expansion histories.
A restricted mathematical-expression parser that avoids the security and reproducibility risks of unrestricted code execution.
Automatic and transparent generation of reproducible Cobaya configuration files.
Isolated execution of MCMC analyses with persistent status and logging.
Integrated posterior inspection and visualization via GetDist.
Support for cosmic chronometer data and official Cobaya Type Ia supernova likelihoods.
Comprehensive validation through synthetic parameter recovery and comparisons with independent Cobaya calculations.

Results and Validation
The reliability of CosmoFit was evaluated through three complementary procedures:

Automated Testing: Unit and integration tests covered the expression parser, configuration builder, likelihood layer, and execution worker. A "smoke test" successfully executed a minimal $\Lambda$ CDM analysis from configuration to posterior output.
Synthetic Parameter Recovery: Using a synthetic cosmic-chronometer dataset generated with fiducial parameters ( $H_0 = 70$ , $\Omega_m = 0.30$ ), the full workflow successfully recovered these values within the expected 68% and 95% credible regions. The recovered $H_0$ was $71.06 \pm 1.58$ km s $^{-1}$ Mpc $^{-1}$ , consistent with the fiducial value.
Independent Comparison: Analyses generated via CosmoFit were compared against independently constructed Cobaya runs using the same flat $\Lambda$ CDM model and cosmic-chronometer data. The posterior means agreed to within less than 1% (e.g., $\delta\mu(H_0) = 0.0026\%$ ). Differences in posterior standard deviations were attributed to expected variations in finite MCMC realizations.

An illustrative application of a spatially flat $w$ CDM model constrained by Pantheon+SH0ES data demonstrated the complete workflow, from $H(z)$ definition to the generation of marginalized constraints and triangle plots.

Significance and Claims
The author positions CosmoFit not as a replacement for established inference engines, but as an accessible, transparent, and reproducible layer that simplifies the configuration and execution of background-level cosmological analyses. The paper claims that the tool effectively reduces the technical barrier for users wishing to test explicit $H(z)$ models without sacrificing the scientific rigor required for research and academic applications. By preserving the underlying statistical problem and ensuring that all configuration files and data products are stored independently, CosmoFit bridges the gap between fully manual pipeline construction and closed "black-box" applications. The validation results support the interpretation of the interface as a reliable orchestration layer over Cobaya, ensuring that graphical convenience does not compromise the integrity of the inference process.

More like this