Identifying IntegralSeparable Dimension Pairs - PowerPoint PPT Presentation
1 / 13
Title:
Identifying IntegralSeparable Dimension Pairs
Description:
Task: Please search cars with low price and high score! Which figure makes it easier? Conclusion: Separable dimensions are more suitable for describing independent ... – PowerPoint PPT presentation
Number of Views:12
Avg rating:3.0/5.0
Title: Identifying IntegralSeparable Dimension Pairs
1
Identifying Integral-Separable Dimension Pairs
- Zaixian Xie
- Mar 15, 2006
2
What are them?
Are BC more similar?
Are AB more similar?
3
What are them?
- Integral display dimensions Two or more
attributes of a visual objects are perceived not
independently. (e.g. x-size and y-size) - Separable display dimensions People tend to make
separable judgments about each graphical
attribute. (size and gray scale)
4
How to use them
a. Integral Dimensions x-size price y-size
score
b. Separable Dimensionssize price gray-scale
score
Task Please search cars with low price and high
score! Which figure makes it easier?
Conclusion Separable dimensions are more
suitable for describing independent multiple data
attributes on glyphs to visualize multivariate
data.
5
Experiment Design
Since x-size and y-size are perceived integrally,
B and C are perceived as more similar.
X and Y are two visual dimensions to test
Since y-size and gray-scale are perceived
independently, A and B are perceived as more
similar because of the same y-size.
6
Experiment Design
- Dimensions to Test
- x-size
- y-size
- color
- gray scale
- orientation
- shape (circle, square)
7
Experiment Design
- Dimension Pairs to Test
- color vs. shape
- orientation vs. color
- orientation vs. gray scale
- x-size vs. color
- x-size vs. gray scale
- x-size vs. orientation
- x-size vs. shape (circle, square)
- y-size vs. x-size
8
Experiment Design
- 23 Participants
- 14 graduate students
- 15 beginners on visualization
9
Experiment Result
10
Experiment Result
- Analysis on the number of questions on which each
user answer a (separable dimensions) - average 4.44
- variance 2.53
11
Conclusion
- We should map more separable dimensions to
multiple data attributes of glyphs - A continuum of integral-separable is more
accurate to present the fact. - Difference exists among subjects.
12
Improvement on Experiment
- More strict environment to control display time.
- Exchange the two dimensions of the pairs.
- Looking for the relationship between culture,
gender and experiment result.
13
Questions or Comments
Recommended
CrystalGraphics Presentations
