Concepts

1. What are geometries?

• Points
• Polygons
• Lines

2. data vs attributes

The type of geometry and the amount of geometires determine the tool stack

• What kind of data do you have?

  • Points

    • How much data?

      • Just enough: Convert the data to GeoJSON & make a simple Leaflet map

      • Too much, but each point's data is important: Cluster your points with Leaflet.markercluster

        

        example: http://open.undp.org/#2014/filter/operating_unit-AFG

      • Too much, the points have some value that can be aggregated: Create hexbins of your points with the QGIS hexbin plugin

        

      • Too much, the points just represent presence - like tweets: Create a heatmap with Leaflet.heat or QGIS heatmap plugin.

        

  • Polygons

    • How much data?

      • Just enough:Convert the data to GeoJSON & make a simple Leaflet map

        

        example: http://leafletjs.com/examples/choropleth.html

      • Too much, the polygons have necessary detail: Use TileMill to render an interactive map with UTFGrid. If you've ever made a map with tilemill, the interactivities are using UTFGrid. See tutorial.

      • Too much, the polygons have unnecessary details: Simplify them with TopoJSON or QGIS

    • What kind of attributes do they have?

      • Absolute numbers:Convert the points to centroids with QGIS
      • Normalized the absolute: Make a choropleth map with Leaflet for small data, TileMillfor big data

• Visualization defaults

  • Projection:

    • If it's a web map with tiles, it's using Spherical Mercator
    • If using d3 and not using tiles anywhere, use whatever fits best. Bonus projections are in d3-geo-projection.

  • Colors:

    • When in doubt, use ColorBrewer
    • Want to know more? Read Subtleties of Color

  • Scales:

    • For any data
      • Try linear first
      • Then quantile
    • For data of rates or compounding values
      • Try log and power scales: "A logarithmic scale is a scale of measurement that displays the value of a physical quantity using intervals corresponding to orders of magnitude, rather than a standard linear scale."

  • Points:

    • Start with normal circles with no strokes
    • Scale points by area, not diameter (proportional)

  • Flair:

    • Only add a north arrow if north isn't up
    • Always attribute your data, especially OpenStreetMap, to avoid the nerd wrath
    • If it zooms, add visible zoom controls. Pan isn't necessary, but not everyone has a scroll wheel / multitouch