{ "metadata": { "name": "" }, "nbformat": 3, "nbformat_minor": 0, "worksheets": [ { "cells": [ { "cell_type": "code", "collapsed": false, "input": [ "%autosave 10" ], "language": "python", "metadata": {}, "outputs": [ { "javascript": [ "IPython.notebook.set_autosave_interval(10000)" ], "metadata": {}, "output_type": "display_data" }, { "output_type": "stream", "stream": "stdout", "text": [ "Autosaving every 10 seconds\n" ] } ], "prompt_number": 1 }, { "cell_type": "markdown", "metadata": {}, "source": [ "http://www.functionalelegance.com\n", "\n", "## History\n", "\n", "- John Snow's cholera investigation; plot dots on a map\n", "\n", "## \"Hockey stick\" chart\n", "\n", "- Famous climate change temperature change chart; controversial\n", " - All down to the axis scale!\n", " - Such a simple choice is a frame; we choose to present a bias, honestly or not.\n", "\n", "## Aspect ratio\n", "\n", "- Another simple question! How do you choose it?\n", "- 1:1 - fair?\n", "- 3:2 - landscape?\n", "- golden ratio - beautiful?\n", "- average slope 45 degrees - perceptually optimal for orientation discrimination.\n", " - easiest angle to see deviations in trend (rather than horizontal / vertical)\n", "- but graphic designers know - all choices depend on what story you want to tell\n", " - choices inevitably affect this.\n", "\n", "##\u00a0Map projections\n", "\n", "- Mercator - preserve angles, not areas.\n", " - used for shipping / navigation\n", "- Choices are frames. You understand \"3D projections onto 2D are necessarily imperfect\", but ordinary people just blindly accept the frame.\n", "- Robinson - almost preserve area, not angle\n", "- But why not plot, scaling for GDP and population, not area? Surely more informative.\n", "\n", "## Choices\n", "\n", "- Representing numeric values without misleading users - our goal.\n", "- Difficult to judge area of quadrangles\n", " - Yet we use quadrangular heat maps!\n", "- **Ebbinghaus Illusion**: very difficult to perceive areas of circles based on context\n", " - But we use bubble charts!\n", "- Even if you use numeric labels, it's too late. Your users have made judgements visually.\n", "- Can't distinguish colours if they're next to other colours\n", " - Yet we use heat maps!\n", "\n", "## Context\n", "\n", "- Add context!\n", " - Textual callouts\n", " - Don't use many colour graduations, few\n", " - Different charts on same page for different views - allows different perspectives, robust.\n", "- The more people are paid, the less able they are to read charts.\n", "- Use familiar idioms\n", " - Next to a vertical bar chart, add traffic lights! Green is up, red is down, yellow is OK.\n", "- \"Familiar visual metaphors make interpretation easier\"\n", "\n", "## Colour gradients\n", "\n", "- Don't use rainbow palettes for continuous numeric values.\n", "- We always see edges between hues (turns categorical).\n", "- Yellow stands out too much\n", " - Use iso-luminate palette, yellow turns brown.\n", "- Brain can distinguish brightness much better than hue.\n", "\n", "## Transparency layering\n", "\n", "- !!AI see presenter's website for info; different ways of mixing layers.\n", "\n", "## Sphere of Influence graphs\n", "\n", "http://demonstrations.wolfram.com/SphereOfInfluenceGraphs/\n", "\n", "## Graphs\n", "\n", "http://visualization.geblogs.com/visualization/network/\n", "\n", "## Summary\n", "\n", "- Since framing is inevitable, start with the user and their context and objecives.\n", "\n", "## Questions\n", "\n", "- Given interaction *or* offer multiple simultaneous views, which to do?\n", " - Interaction always wins.\n", " - Multiple charts will often leave users confused." ] } ], "metadata": {} } ] }