Friday 28 November 2025
Scientists have developed a new method for assessing the reliability of data visualizations, which could help prevent misinterpretation and misuse of complex datasets.
When it comes to understanding high-dimensional data, such as that obtained from gamma-ray bursts or other astrophysical phenomena, scientists often rely on dimensionality reduction techniques like UMAP (Uniform Manifold Approximation and Projection) to visualize the data in a lower-dimensional space. However, these methods can be prone to errors, particularly when it comes to identifying trustworthy patterns and relationships.
To address this issue, researchers have developed an innovative approach that combines local neighborhood preservation with statistical analysis to evaluate the trustworthiness of individual data points in the reduced-dimensionality space. The method, known as scDEED (Statistical Confidence for Dimensionality-reduced Embeddings), uses a combination of Pearson correlation and permutation tests to assess the reliability of each point’s position in the 2D embedding.
The approach is based on the idea that if a data point is truly representative of its neighborhood in the original high-dimensional space, it should also be well-preserved in the reduced-dimensionality space. By comparing the distances between points in both spaces, scientists can evaluate how well the neighborhood structure is preserved and assign a reliability score to each point.
The method was tested on a dataset of gamma-ray bursts, which are brief, intense explosions that occur when massive stars collapse or merge with other objects. The researchers found that around 90% of the events were classified as trustworthy, while only a small number (5 events) were deemed dubious. The remaining points fell into an intermediate category, where the reliability score was ambiguous.
The results have significant implications for data analysis and visualization in astrophysics and beyond. By using this method, scientists can gain more confidence in their findings and avoid misinterpreting complex datasets. The approach could also be applied to other fields, such as medicine or finance, where high-dimensional data is increasingly common.
While there are still limitations to the method, particularly when it comes to identifying subtle patterns and relationships, the researchers believe that scDEED offers a valuable tool for assessing the trustworthiness of reduced-dimensionality embeddings. As scientists continue to push the boundaries of data analysis and visualization, this innovative approach could play an important role in ensuring the accuracy and reliability of their findings.
Cite this article: “Evaluating the Trustworthiness of Data Visualizations”, The Science Archive, 2025.
Data Visualization, Reliability Assessment, Dimensionality Reduction, Umap, Scdeed, Statistical Analysis, Pearson Correlation, Permutation Tests, Neighborhood Preservation, Astrophysics
