Data flow and Visualization details
The Geographic Analytics Extension provides many visualizations. Each visualization style listed below is one of the drop-down options found in the Configure Map Appearance menu. This section will explain how the extension processes data and uses that data to create visualizations.
Data flow
The Geographic Analytics Extension works by tapping into the data that powers your Tableau worksheets. This data is temporarily cached for downstream processing, but is overwritten on a new selection or a visualization update.
Once connected, it utilizes a library called Turf to calculate the bounding box of the selected data. This is recalculated on every worksheet interaction, which is how it produces the zoom-to-selection function.
Visualization Types
Clustering
Clustering is the default visualization. This works by passing the data into Mapbox GL and requesting a style-type of cluster. This cluster is powered by a library called Supercluster that figures out how close points are to each other within a certain radius and then creates a single point (the cluster) out of those points. The more points, the larger the cluster. This is computed for all points as you move around the map. As you zoom in, the clusters disassemble into their constituent points. You can read more about Supercluster on the Mapbox blog and on the Supercluster GitHub repository.
Grids
Grids are calculated using a Turf method called hexGrid and squareGrid. This takes the bounding box of the data (collected when the Geographic Analytics Extension connects to the Worksheet) and a distance setting (in miles or kilometers), and outputs a grid covering that area.
The extension then uses a secondary method called geometryCollection. This takes the grids that it generates and compares it to the points from the worksheet. Once the extension has collected the points that are within each cell, it then computes the Measures and assigns them to each cell.
Grids can also be extruded, which takes the value that is assigned to the color and moves the grid into 3D. This is where the Extrusion Height is used, as you can adjust the minimum height so that your data is visible and variance between cells can be observed. To trigger extrusion, click the Awesome button. To reset, click the Reset button.
Voronoi
Voronoi are calculated like Grids - they take in data and pass it through a Turf method called voronoi. Since each Voronoi cell contains one data point - there is no collection, color is the selected measure assigned to that point.
Data elevation
Data elevation, in GIS terms, is an isoband. Isobands are traditionally used to show changes in data, typically elevation, in two dimensions. They are computed from a grid that expresses how data changes over space which then generates rings that show a specific value and rate of change. The closer the rings, the faster the rate of change; the further, the slower.
This is an intensive calculation that may cause the extension to become unresponsive, depending on your settings.
Creating an isoband is a two step process. First, the extension will take the data from the worksheet and process it using a Turf method called interpolate. This takes the points from the worksheet and snaps them to a grid of width X, which is the width you established in the configuration screen. This process will interpolate the change in your selected measure via a process called inverse distance weighting.
Remember: Before creating a data elevation, consider the fact that all computations are happening within a browser. The tighter the grid, the longer it will take to process. The method has to do the interpolation and then snap points to the grid, which can be intensive if the grid is too tight over too large a geographic area.
Second, the grid must be processed into bands representing areas of change and when breaks are exceeded. This is performed by a Turf method called isobands. It takes the grid from the first step and then finds the breaks in the data where changes occur. These steps are controlled by your Classification method and the number of breaks you select.
For example: if your data range is 0-100, and you have four equidistant breaks, you will see three lines that show where you cross 25, 50, and 75 (calculated from the interpolated grid).
Before choosing Data Elevation, you may want to switch to a Square or Hex grid to see how tightly packed your grid might be. Once you have found a grid that works, experiment with your breaks. Once you have completed both of those steps, switch to Data Elevation and it should do well.