Overview
Searching for, identifying, and validating the shape of experimental datasets is an important component of scientific data analysis. Typically, these shape-based tasks are implemented manually or using broadly heuristic methods. We have developed a Bayesian method to objectively calculate the intrinsic shape of data using probability theory. This general method uses Bayes’ rule to select the best shape for a specific dataset given the uncertainty present in the estimation due to noise and other nuisance parameters. By providing analytical formulas to determine the shape of data, we hope to enable a machine learning-based approach to analyses that were previously done heuristically or “by eye”. This method is broad, readily generalizable, and can be applied to many techniques in the physical and life sciences. Check out our preprint and example gallery to see if this method has potential applications in your research project!
This project was initially developed as a collaboration between the Kinz-Thompson and Gonzalez labs. All code is open-source and free to use. We hope to encourage further implementations and welcome collaborations to extend Bayesian shape calculation to different fields of the physical and life sciences.
This website serves as a home-page for all applications of this Bayesian shape calculation method. As particular implementations are developed, they will be aggregated here. Please contact us if you have an application that you would like added to this page, or if you are interested in creating one!