Cours de Visualisation d'information InfoVis Lecture. Multivariate Data Sets

Cours de Visualisation d'Information

InfoVis Lecture

Multivariate Data Sets

Frédéric Vernier

Maître de conférence / Lecturer Univ. Paris Sud

Data Sets

Ø

Data comes in many different forms

Ø

Typically, not in the way you want it

Ø

How is stored (in the raw)?

Ø

Heterogeneous

data often seen as

multiple dimensions of elements extracted

by patterns or needs.

Schema

Ø

Cars

Ø

brand

Ø

model

Ø

year

Ø

cost

Ø

size

Ø

weights

Ø

miles per gallon

Data Tables

Ø

Often, we take raw data and transform it

into a form that is more workable

Ø

Main idea:

Ø

Individual items are called

cases

Variable Types

Ø

N

-Nominal (equal or not equal to other values)

Ø

Example: gender, hair color

(blond, brown, black, red)

Ø

O

-Ordinal (obeys < relation, ordered set)

Ø

Example: soccer leagues, rainbow colors

Ø

Q

-Quantitative (can do math on them)

Variable Types

Ø

Three main types of variables

Ø

N

-Nominal

Ø

By Class: data belong or not to classes (.org, .com, .fr)

Ø

Partially ordered: order on classes (engineer students)

Ø

O

-Ordinal

Ø

Q

-Quantitative

Ø

Quantitative + 0 (clear 0)

Example

Baseball

statistics

Metadata

Ø

Descriptive information about the data

Ø

Might be something as simple as the type of a

variable, or could be more complex (INT)

Ø

For times when the table itself just isn’t enoughi

Ø

AtBats

≥

Hit

≥

HomeRuns

Ø

if “YearInMasterLeague”=1 then AtBats=CareerAtBat

1 M2R InfoVis Lecture. 2011. Univ. Paris Sud

How Many Variables?

Ø

Data sets of dimensions 1,2,3 are common

Ø

Number of variables per class

Ø

1 - Univariate data (e.g timeline)

Ø

2 - Bivariate data (e.g maps)

Ø

3 - Trivariate data (volume)

Ø

>3 - Hypervariate data (???)

Ø

Example:

www.nationmaster.com

Univariate

Ø

Representations

Ø

Dot plot

Ø

Bar chart (item vs. attribute)

Ø

Tukey box plot

Ø

Histogram

5 3

Bivariate

Ø

Scatterplot

Common

BUT

Trivariate

Ø

3D scatterplot, 2D plot+size

Hypervariate Data

Ø

What about data sets with

MANY

variables?

Ø

Often the interesting ones

Ø

n-D

What does 10-D

space look like?

Multiple Projections

Give each variable its own display

A B C D E

1 4 1 8 3 5

2 6 3 4 2 1

3 5 7 2 4 3

4 2 6 3 1 5

A B C D E

1

2

3

4

Help me Infovis !

Ø

smart layout

Scatterplot Matrix

All pair of variables in

their own 2-D scatterplot

Brushing (subset)

&

Linking (sync.)

label, dot plot, scale

Histogram

>

dot plot

for distribution

Scale

row &

column

Chernoff Faces

Simple

Example

[Spinelli and Zhou, 2004]

On steroids

Star Plots / Glyphs

Var 1

Var 2

Var 3

Var 4

Var 5

Value

Space out the n

variables at equal

angles around a

circle

Each

“

spoke

”

encodes

examples

circular

// coords

On prednizone ...

just

2 dims

[bertillon]

population

x

percent foreigners

Star Coordinates

E. Kandogan, “Star Coordinates: A Multi-dimensional Visualization Technique with Uniform Treatment of Dimensions”, InfoVis 2000

1 M2R InfoVis Lecture. 2011. Univ. Paris Sud

Demo - Interaction

Ø

Activate/ deactivate axis

Ø

Color selection or axis

Ø

Glyph coordinates

Ø

Scale axis

Ø

Rotate axis

Ø

Dot size

Ø

Brushing on axis

Ø

Trail

Ø

Inspector

Ø

Panning

Parallel Coordinates

V1 V2 V3 V4 V5

By A. Inselberg

Encode variables along

a horizontal row

Vertical line specifies

values

Parallel Coords Example

Basic

Grayscale

Color

From: Dean F. Jerding and John T. Stasko

VisDB

Ø

Database of data items, each of n

dimensions

Ø

Issue a query that specifies a target value

of the dimensions

Ø

Often get back no exact matches

Ø

Want to find near matches

Ø

Relevance factor

Taken from:

Technique

Ø

Calculate relevance of all data points

Ø

Sort items based on relevance

Ø

Use spiral technique to order the values

Ø

Color items based on relevance

Display Methodology

Dim 1

Dim 2

Spiral

in each

window

Items ordered by total relevance

Same item

appears in

same place

in each

window

Highest relevance

value in center,

decreasing values

grow outward

Alternative

Ø

Grouping arrangement => single window

Ø

Create all relevance dimensional

depictions for an item and group them

Ø

Spiral out the

different data

items

Example

Multi-window

Grouping

8 dimensions

1000 items

Overview

More techniques ?

Ø

Combinations

Ø

More integrated software

Highlighted Dynamic Table

Viewer

Nada Golmie &

Bill Kules

Eureka / TableLens

Rao &

Card 94

EZChooser:

Comparisons

Ø

ParCood: <1000 items, <20 attrs

Ø

Relate between adjacent attr pairs

Ø

StarCoord: <1,000,000 items, <20 attrs

Ø

Interaction intensive

Ø

TableLens: similar to par-coords

Ø

more items with aggregation

Ø

Relate 1:m attrs (sorting), short learn time

Ø

Visdb: 100,000 items with 10 attrs

MultiVariate Visu Tools

Paper presentations

Ø

Hajar Falih

Ø

Multi-Dimensional Detective

Ø

Thibaut Jacob

Ø

Rolling the Dice: Multidimensional Visual

Exploration using Scatterplot Matrix Navigation

90 min Lecture: Multi-dimensional Data Visualization Δ