Implementation of Textile Plot

1
Implementation of Textile Plot

• Natsuhiko KUMASAKA
• Fundamental science and technology
• Keio university JAPAN
• Ritei SHIBATA
• Department of mathematics
• Keio university JAPAN

2
Overview

• Textile Plot
• - High dimensional data visualisation
• - Modified parallel coordinate plot
• - Any data type
• Textile Plot on R
• - Part of DandDR
• 54-dimensional data example

3
(No Transcript)
4
Petal.Length
Petal.Width
Sepal.Width
Sepal.Length
5
virginica
versicolor
setosa
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
Criterion locations and scales

• Data Vectors
• Coordinate vectors
• Location parameter vector
• Scale parameter vector
• Ideal coordinate vector

11
Categorical data

• Coordinates of levels are automatically chosen

Using a set of contrasts
Coordinate vector
12
Ordered categorical data

• Order of levels is retained

Using a specific contrast
or
Coordinate vector
13
Missing values

• Indicator matrix

14
Order of axes Variance criterion

• Variance of the coordinate vector
• Further left axis has bigger variance

The further left axis is closer to the ideal
coordinate vector.
15
Order of axes Axis clustering

• Linkage method
• Ordered single end-linkage(Hurley 2004)
• Distance between variables
• Sum of absolute values of slopes

16
Display design

• Numerical data
• Continuous data
• Continuous line
• Discrete data
• Tick marks
• Arrow head to show the orientation
• Possible minimum and maximum
• Non-numerical data
• Possible levels
• Ordered categorical data
• Arrows
• Logical
• Coloured
• All data
• Multiplicity on the coordinate is represented by
  the area of the circle
• Missing value
• Label (with unit or numeral)

17
Implementation of Textile Plot

• DandDR (http//www.stat.math.keio.ac.jp/DandDIV/)
• -Add-on package for R
• -Interface between DandD (yokouchi and Shibata
  2004) and R
• -Receiving data and necessary attribtues
• -Creating a dad object on R
• -List object which consists of data and
  attributes
• -Textile plot implemented as plot method
• Computation
• C Language
• -Matrix computation
• -Optimasation problem (cLapack)
• Display
• R graphical functions (points, segments, etc)

18
Scalability Computation of

• Test data
• (Random numbers)
• Condition
• -Xeon CPU 3.2GHz
• -2 GB RAM

19
Body Measurement (Kouchi et al. 2000)

• Foot circumference
• Foot length
• Forearm circumference
• Forearm length
• Hand breadth
• Hand length from crease
• Hand length from stylion
• Hand thickness
• Heel breadth
• Hip circumference
• Instep length
• Lateral epicondyle height
• Lateral malleolus height
• Maximum body height
• Medial malleolus height
• Subscapular skinfold thickness
• Suprailiac skinfold thickness
• Suprasternal height
• Symphyseal height

76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
92 93 94 95 96 97 98 99 100 101 102 103 104 105 10
6
Item Sex Age Educational background Occupation Rac
e Body mass Stature Iliac spine height
standing Shoulder biacromial breadth Head
length Head breadth Chest circumference Waist
circumference Calf circumference Ball angle Ball
breadth Bicondylar femur Bicondylar
humerus Bicristal breadth Calf skinfold
thickness Cristal height Fibular instep
length Foot breadth
ID 2 6 8 9 10 11 12 16 30 48 49 62 63 66 67 68 69
70 71 72 73 74 75
54 Variables
20
(No Transcript)
21
No variation Educational background
Occupation Race
22
(No Transcript)
23
Cluster 1
Cluster 2
Cluster 3
24
Cluster 1
25
72 Calf skinfold thickness 99 Triceps skinfold
thickness 92 Suprailiac skinfold thickness 91
Subscapular skinfold thickness
Cluster 1
26
83 Hand thickness 80 Hand breadth 70 Bicondylar
humerus Sex 88 Lateral malleolus height 90
Medial malleolus height 30 Shoulder breadth
Cluster 1
27
83 Hand thickness 80 Hand breadth 70 Bicondylar
humerus Sex 88 Lateral malleolus height 90
Medial malleolus height 30 Shoulder breadth
Cluster 1
Female
Male
28
Cluster 1
Female
Male
29
Cluster 2
30
Cluster 2
31
Cluster 2
103 Upper arm length 104 Upper limb length 79
Forearm length
32
89 Maximum body height 12 Stature 93
Suprasternal height 87 Lateral
epicondyle height 94 Symphyseal height
16 Iliac spine height standing 100
Trochanterion height 73 Cristal height 106
Waist height
Cluster 2
33
74 Fibular instep length 77 Foot length 86 Instep
length 81 Hand length from crease 82 Hand length
from stylion
Cluster 2
34
Cluster 2
Arm
Leg and Body
Foot
Hand
35
Cluster 3
36
Cluster 3
37
105 Waist breadth 63 Waist circumference 62
Chest circumference 101 Upper arm
circumference 102 Upper arm circumference
flexed 78 Forearm circumference 11 Body mass
38
Cluster 3
66 Calf circumference 95 Thigh circumference 85
Hip circumference 71 Bicristal breadth
Age 97 Toe V angle 67 Ball angle 96 Toe I
angle 49 Head breadth 98 Total head height 48
Head length
39
Cluster 3
66 Calf circumference 95 Thigh circumference 85
Hip circumference 71 Bicristal breadth
Age 97 Toe V angle 67 Ball angle 96 Toe I
angle 49 Head breadth 98 Total head height 48
Head length
knot
40
Cluster 3
84 Heel breadth 76 Foot circumference 75 Foot
breadth 68 Ball breadth 69 Bicondylar femur
41
Summary of body measurement data

• Cluster1
• Sex Hand and Shoulder
• Skinfold thickness
• Cluster2
• Height Arm and leg lengths
• Hand and foot lengths
• Cluster3
• Weight Upper body size
• Lower body size, foot
  shape
• and head
  size

42
Bibliography

• Hurley, Catherine B., Clustering Visualisations
  of Multidimensional Data, Journal of
  Computational and Graphical Statistics, 13 (2004)
  788--806.
• Kouchi, M. , Mochimaru, M., Iwasawa, H., Mitani,
  S., Anthropometric database for Japanese
  Population1997-98, Japanese Industrial Standards
  Center (AIST, MITI) (2000).
• Kumasaka, N. and Shibata, R., Implementation of
  Textile Plot, Proceedings of the COMPSTAT 06
  (2006).
• LAPACK Home Page, http//www.netlib.org/lapack/
• Yokouchi, D. and Shibata, R., DandD Client
  Server System, Proceedings of the COMPSTAT 04
  (2004).