glue: Linked-View 
Exploratory Visualization of 
High-Dimensional Data, 
for Everyone

Alyssa Goodman's single slide for lightning session at NSF SI2 PI meeting April-May 2018, introducing poster at https://figshare.com/s/7aacc37dc44a0e410587.<div><br></div><div><br></div><div><b>ABSTRACT</b> of NSF SI2-SSE-1739657 & 1740229, entitled <i>"Collaborative Research: A sustainable future for the glue multi-dimensional linked data visualization package"</i><div><div><br></div><div>Glue is a free and open-source application that allows scientists and data scientists to explore relationships within and across related datasets. Glue makes it easy create a wide variety of visualizations (such as scatter plots, bar charts, images) of data, including three dimensional views. What makes Glue unique is its ability to connect datasets together, without merging them into one. Thus, for example, two Earth-based mapping data sets may be connected and jointly visualized by using the coordinates (e.g. latitude and longitude) to glue the maps together, so that when a user selects (e.g. with a lasso tool) regions in one data set, the corresponding selected subset of data will highlight in all related visualizations simultaneously. These "linked views" are especially powerful across wide varieties of plot types. For example, if a user interested in air traffic control glues a data set with information about the 3D locations of all airplanes to a second data set giving weather information, that user could make a combination of selections that would highlight (on maps, in 3D views, or any other display) planes at particular altitudes where thunderstorms might be likely to occur within a specific period of time. In particular, Glue makes it easy for users to create their own kinds of visualizations, which is important because different disciplines often need very specialized ways of looking at data. The software is already being used widely across several disciplines, in particular, astronomy and medicine, for which has been specially optimized. This project is adding new features to make Glue more useful in more fields of science (e.g. bioinformatics, epidemiology) where there is demand for linked-view visualization, as well as making it more accessible as an educational tool. In addition, this project is training new users and developers, who will expand Glue into a much more sustainable community effort. <br><br>Glue is an open-source package that allows scientists to explore relationships within and across related datasets, by making it easy for them to make multi-dimensional linked visualizations of datasets, select subsets of data interactively or programmatically in 1, 2, or 3 dimensions, and see those selections propagate live across all open visualizations of the data (e.g. graphs, maps, diagnostics charts). A unique feature of glue is that datasets from different sources can be linked to each other, using user-defined mathematical relationships between sets of data components, which makes it possible to carry out selections across datasets. Glue, written in Python, is designed from the ground-up for multidisciplinary work, and it is currently helping researchers make discoveries in geoscience, genomics, astronomy, and medicine. It is also giving insights into data from outside academia, including open data provided by governments and cities. To become sustainable in the long term, glue development is a community-driven effort. Through tutorial and developer workshops, coding sprints, and strategic collaborations with researchers in several disciplines and experienced open source developers, the glue team is helping user communities extend glue by developing new functionality useful within particular fields of research. The team is helping users contribute the most widely-needed functionality back to glue, and is recruiting active contributors to participate in core glue development. As the community grows, glue development is being guided to focus on several major features useful to the broad research community, including: support for very large datasets, support for running glue fully in the browser (inside Jupyter notebooks and Jupyter Lab), and improved interoperability with third-party tools.</div></div></div>