Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 1Niklas Elmqvist | Purdue UniversityPierre Dragicevic | INRIAJean-Daniel Fekete | INRIA

“ ”Rolling the Dice: Multidimensional Visual Exploration

using Scatterplot Matrix Navigation

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 2

Visualizing Complex Data

Complex dataset

Complex visualization

Many simple visualizations

?

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 3

Motivation Information visualization is now more than

15 years old Lots of visual representations

New ones every year at IEEE InfoVis Varying complexity Varying purpose

Scatterplots: one of the simplest and most widely used visual representations Multidimensional data Examples: SpotFire, ADVIZOR, etc

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 4

Scatterplots

Assign dimensions to graphical axes Two (typical) or three

Data cases as points in the space defined by axes

Additional dimensions Point color Point size Point shape …

Limited number of displayed dimensions! dimension A

dimension B

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 5

Problem Many (most?) interesting real-world

datasets have many dimensions Certainly more than a scatterplot can show!

Two possible solutions:1. More complex visual representation

Parallel coordinates, DataMeadow, etc… Lose simplicity of scatterplots

2. Create many scatterplots (one per combination of dimensions) But how to visualize them?

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 6

Scatterplot Matrices Idea: Create matrix of all data dimensions

Each column/row is a data dimension Each cell in the matrix is a scatter plot Reordering of columns/rows (automatic or manual)

Explored by others, but we add a twist: Use matrix as a space for navigation Visual exploration becomes a navigation problem

Result: Visualize complex data through sequence of simple visualizations

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 7

Example: Scatterplot Matrix

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 8

Navigating Scatterplots Idea: Navigating from scatterplot A to

scatterplot B using an animated transition Problem with linear interpolation animation

No semantic meaning to the user Can be complex to follow

Compounded by large datasets Can we add meaning to the transition?

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 9

Adding Meaning to Transitions Make the transition between scatterplots in the

scatterplot matrix easier to follow: Restrict to rectilinear movement (no diagonal

movement) Change only one visualized data dimension at a time

Utilize unused third graphical dimension for the new dimension to show during the transition

Metaphor: rotating a 3D scatterplot to show new 2D projection Like rolling dice to see another side

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 10

Using the Third Dimension Lazy allocation 3rd graphical dimension

Invisible when viewing a 2D scatterplot, so can be assigned any data dimension

Allocation performed only during the transition Not “true” 3D – we use third dimension as a

presentation tool only Mechanics of performing 3D rotation:

Perspective view: 3-stage animation (extrusion, rotation, projection) Looks more natural, but requires three steps

Orthographic view: 1-stage animation (rotation) Only one step, but may look “weird”

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 11

Example: 3D Perspective Transition

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 12

Putting It Together ScatterDice application

Built in Java and uses OpenGL through JOGL Loads multidimensional datasets using the

InfoVis Toolkit [Fekete 2004] Exploration by navigating a scatterplot

matrix 3D transitions between adjacent scatter plots

Support reordering of dimensions Automatic: by correlation between

dimensions Manual: drag and drop of rows columns

Interaction for the navigation is important

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 13

Interaction Techniques

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 14

Demonstration

DEMO

Rolling the Dice: Multidimensional Visual Exploration using Scatterplot Matrix Navigation 15

Conclusions and Future Work Idea: Combine multiple simple visualizations

using interactive animation Visualize complex datasets Provide meaningful transitions

Interaction is a key element Avoid automated tours Allow refinement of queries

Future work Empirical evaluation of this method Other applications of the same approach?

Questions? Niklas Elmqvist

Electrical & Computer EngineeringPurdue UniversityWest Lafayette, IN 47907-2035E-mail: elm@purdue.eduWeb: http://engineering.purdue.edu/~elm/