H517 Visualization Design, Analysis, & Evaluation Week 6: - PowerPoint PPT Presentation

H517 Visualization Design, Analysis, & Evaluation Week 6: Marks & Channels (cont’d) Tables and multi-dimensional data Khairi Reda | redak@iu.edu School of Informa5cs & Compu5ng, IUPUI

Administrativia • Coming up next week: Project 1 presentations • 4 min presentation each + 45 sec Q&A ( sharp limit! ) • No need to prepare PowerPoint slide, just bring up your vis and show it to class • You’ll not be allowed to use your own laptop; need to be able to access the vis through a public URL • Demo the vis, talk about your design process, challenges encountered and how you addressed them • Audience: ask question, give feedback, critique the vis; always provide constructive comments

Last week…

Marks Graphical elements in an image points (0D) lines (1D) areas (2D) volume clouds (3D)

Channels (aka Visual Variables) Parameters that control the appearance of marks based on a;ributes

magnitude channels iden=ty channels good for ordered attributes good for categorical attributes Tamara Munzner Via Miriah Meyer

How much longer? 4x Alex Lex

How much larger (area)? 5x Alex Lex

Psychophysics Steven’s Psychological power law perceived sensation S = I n physical intensity

Tamara Munzner Via Miriah Meyer

Heer & Bostock, 2010

Discriminability can channel differences be discerned? Via Miriah Meyer

Position Offers very good discriminability posi5on

Position But this doesn’t extend to 3D! Perspec5ve distor5on Occlusions

Factors affecting accuracy of Length/Position judgement aligned stacked bar chart unaligned (unaligned)

Separable vs Integral channels separable channels: can be judged individually integral channels: are viewed holis5cally separable integral Ware 2004 Based on a slide by Miriah Meyer

Chernoff faces

Chernoff faces

This week Tables and multi-variate data

Key attribute

Visualizing Tables

Visualizing Tables 1 quan=ta=ve a;ribute 1 categorical (key) 2 quan=ta=ve a;ributes

Visualizing Tables key attribute key attribute

Don’t use line charts for categorical a;ributes! ok: “Men are taller than bad: “The more male a women (on average)” person is, the taller he is” ok: “Twelve year olds are ok: “The older a person taller than ten years old” the taller he/she is” Miriah Meyer

Arrange Tables Mul5ple columns/categories Streit & Gehlenborg, PoV, Nature Methods, 2014 Via Alex Lex

Arrange Tables

2 quan=ta=ve a;ributes Y X Z Z X Y What if we have and want to see more than 2 quantitative attributes at the same time?

Wilkinson et al., 2005 Via Miriah Meyer nine characteristics of Abalone (sea snails)

Wilkinson et al., 2005 Via Miriah Meyer

Wilkinson et al., 2005 Via Miriah Meyer

Parallel Coordinates V1 V2 V3 V4 V5 10 8 6 4 2 0 Example by Miriah Meyer

Parallel Coordinates posi=ve correla=on straight lines nega=ve correla=on all lines cross at a single point Wegman 1990 Via Miriah Meyer

Parallel Coordinates ProtoVis Via Miriah Meyer

Do you see any correlation? Correla=ons only visible between neighboring axis pairs: axis order ma^ers allow user to reorder axis Fua 1999 Via Miriah Meyer

Hierarchical Parallel Coordinates Fua 1999

Hierarchical Parallel Coordinates Instead of showing all points, show a band represen=ng a cluster: mean: opaque line min/max: illustrated by band width with decreasing opacity from mean cluster Fua 1999

Hierarchical Parallel Coordinates Cluster: lines that share similar shapes. Interac5vely varying the similarity threshold allows us to “unpack” clusters Fua 1999

Radial Layout Star Plot Similar to parallel coordinates, but axes radiate from a common origin Scotch Whiskies Via Alex Lex

Arrange Tables

Arrange Tables Table as a heatmap 1 2 5 4 5 0 0 1 5 6 1 2 2 1 3 1 4 1 2 1

Arrange Tables Table as a heatmap 1 2 5 4 5 0 0 1 5 6 1 2 2 1 3 1 4 1 2 1

Arrange Tables Table as a heatmap 0 0

Arrange Tables Table as a heatmap Order is important: Clustering is o`en used with heatmaps

Next week Project 1 presenta5ons