The craft of Information Visualization

1
The craft of Information Visualization

• NCRM Research Methods Festival 2008
• Jonathan C. Roberts
• School of Computer Science
• Bangor University

2
Minards plot

• http//www.math.yorku.ca/SCS/Gallery/re-minard.htm
  l

3
The 1854 London Cholera Epidemic.
4
Advantages of Information Visualization

• Visualization provides
• The ability to comprehend huge amounts of
  information
• The perception of emergent properties that were
  not anticipated
• problems with the data to be made apparent (e.g.
  errors or artefacts of the data)
• Large/Small scale features can be seen
• facilitation of hypothesis formation

5
Schematic of the visualization process
6
Things to consider

• Six important aspects of an Information
  Visualization
• Data
• Visual Structures
• Multiple Views
• Interaction Exploration
• Tasks ( Management of tasks)
• Level organization

7
1. Data Visual Structures..

• maps interesting data items to graphics objects
• Bertin methodology
• maps the CONTENT (information to be transmitted -
  filtered data) to the CONTAINER (the properties
  of the display/graphic system) using a COMPONENT
  analysis.

8
Bertin COMPONENT analysis

• Bertins component analysis
• invariant and variational components
• number of Components
• length of Components
• organisation of Components

9
Bertin CONTAINER - graphic system properties
Main retinal Variables Position Size Colour
(Hue, saturation, value) Orientation Shape Texture
Additional retinal variables Motion
velocity Motion direction Flicker
frequency Flicker phase

• Representation Styles
• diagrams, networks, maps, symbols
• Retinal Variables
• Level of organisation
• point, line, area, volume

10
Different Mappings
Independent and DependentWhen an experiment is
conducted, some variables are manipulated by the
experimenter (these are called independent
variables) and others are measured from the
subjects (these are dependent variables or
dependent measures.

• 2 variables

dependent
independent
11
Different Mappings
The values are extra dependent values on the same
independent parameter.

• 3 ..4 variables

dependent
independent
12
The data table (spreadsheet)

• This is ok when there is only one independent
  variable. But what if we have multiple
  independents?

dependent
dependent
dependent
independent
13
2D .. 3D
14
Multivariate, Car

• Variable Car1 Car2
• MPG 32 43
• Weight 1000kg 1100kg
• Top Speed 130 140
• 0-60 4 5
• Cylinders 8 6

15
Scatter Plot Matrices
Scatter Plot Matrices
Reorderable matrix
16
Parallel-coordinates (PC or -coords)

• Parallel coordinates yield graphical
  representations of multi-dimensional relations
  rather than just finite points sets.
• Place the axis parallel and join the dots
• Euclidean 3d geometry. X,y,z coordinates
• Point in space is given by extents along the axis
• -coordinates. Point is a line

17
So what is a point

• A n-d point is equivalent to a line in
  -coordinates

http//catt.bus.okstate.edu/jones98/parallel.html
18
Point line duality
l
The line is represented by the crossing
Line in Euclidean
19
Cubes..

• Parallel coordinates provides a very simple
  representation of high dimensional objects such
  as hypercubes. 
• Consider the Parallel coordinate plot of the four
  corners of a two-dimensional square

20
Interacting with - Coordinates
http//software.fujitsu.com/en/symfoware/visualmin
er/vmpcddemo.pdf
21
Selecting a range of records
22
Selecting records
23
Verifying a hypothesis
24
Highlighting relationships
25
Separating different record groups
26
Another observation
27
Visual Structures - Techniques
Graphical properties placing appropriate
marks Substitute different properties with
different marks Aligning data on different
axes composing data Overlaying data on top
28
Multiple Views

• Display different information in different views

Same color
cdv - Cartographic Visualization for Enumerated
Data Dykes
Waltz, Roberts
29
Dual views focuscontext
Table Lens
Dual views Roberts
30
Multiple View Techniques
Different views may be better at displaying that
information Correlations between views can be
highlighted Through brushing or zooming One
view can be for Focus another for context
(focuscontext) One view can be for Overview
another for detail (overviewdetail) Distortion
can be used to (say) place more information in a
small area
31
4. Interaction Exploration

• Allow the user to change their mind and explore
  the data
• To provide sliders/buttons/menus to choose how
  the data is to be viewed
• To select a subset of the information (zoom into
  this)
• E.g. Brushing
• a collection of techniques to dynamically query
  and directly select elements on the visual
  display.
• Usually in dual views (or more)
• Such interaction allows the user to explorethe
  visualization to interactively select a subset
  of points and see how these changesare updated
  in other related views.

32
Zoom

• To focus, Select (or highlight) a feature set of
  information
• Zoom telephoto-lens, reduced field of view
• 3D clipping
• Semantic zoom

Alternate Representations Roberts, Ryan
33
Dynamic queries

• Instant update
• Direct manipulation
• Sliders/buttons

Example of a dynamic queries environment created
with IVEE Measurements of heavy metals in Sweden
FilmFinder Ahlberg, Shneiderman
34
Interaction Techniques
Dynamic Queries (indirect manipulation) Direct
Manipulation Overlays (e.g. magic
lens) Coordination of views which are
coordinated? how are they coordinated?
35
Filter Extract
1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1

• Visual extraction
• constant quantity of information
• brush and highlight
• visually altered to stand out (colour, size ...)
• sliders (1 lt highlight lt 4 ...)
• Subset (filter) of the data
• extract portions of the dataset
• Specialize
• semi-automatic/manual (seed-point, selection)
• neighborhood / global operations

1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1
1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1
36
Filter Extract
1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1

• Visual extraction
• constant quantity of information
• brush and highlight
• visually altered to stand out (colour, size ...)
• sliders (1 lt highlight lt 4 ...)
• Subset (filter) of the data
• extract portions of the dataset
• Specialize
• semi-automatic/manual (seed-point, selection)
• neighborhood / global operations

1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1
1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1
37
Filter Extract
1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1

• Visual extraction
• constant quantity of information
• brush and highlight
• visually altered to stand out (colour, size ...)
• sliders (1 lt highlight lt 4 ...)
• Subset (filter) of the data
• extract portions of the dataset (isolate)
• Specialize/Generalize
• semi-automatic/manual (seed-point, selection)
• neighborhood / global operations

1 1 1 1 1 1 2 3 2 1 1 3 9 3 1 1 2 3 2 1 1 1 1 1 1
3 3 3 3
38
5. Tasks ( Management of tasks)

• Foraging for data
• Solving problems and investigating hypothesis
• Searching for some data (or the lack of data)
• Making quantitative/qualitative analysis
• Querying and finding evidence for decision making

39
Techniques to perform the Task
Overview Zoom Filter Details on
demand Browse Search Read (facts or
patterns) Compare Manipulate Explore Create Dissem
inate and present
From. Readings in information visualization -
Card/Mackinlay
40
6. Level organization

• What is the right level-of-detail?
• Are there too many points on display
  (abstract/summarize/bin/aggregate)
• How is the information organized?
• Think what is close and what is near
• Near objects are easier to compare
• E.g. re-order the axes on a -coord plot

41
Techniques for Level
Delete Re-order Cluster Class Promote Average Abst
ract/Summarize Instantiate Extract Compose Organiz
e
42
Things to remember

• Six important aspects of an Information
  Visualization
• Data
• Visual Structures
• Multiple Views
• Interaction Exploration
• Tasks ( Management of tasks)
• Level organization

43
The craft of Information Visualization

• Jonathan C. Roberts
• School of Computer Science
• Bangor University

END