Brushing Moments in Interactive Visual Analysis Johannes Kehrer 1 , - - PowerPoint PPT Presentation

brushing moments in interactive visual analysis
SMART_READER_LITE
LIVE PREVIEW

Brushing Moments in Interactive Visual Analysis Johannes Kehrer 1 , - - PowerPoint PPT Presentation

Brushing Moments in Interactive Visual Analysis Johannes Kehrer 1 , Peter Filzmoser 2 , and Helwig Hauser 1 1 Department of Informatics, University of Bergen, Norway 2 Department of Statistics and Probability Theory, Vienna University of


slide-1
SLIDE 1

Brushing Moments in Interactive Visual Analysis

Johannes Kehrer1, Peter Filzmoser 2, and Helwig Hauser1

1 Department of Informatics, University of Bergen, Norway 2 Department of Statistics and Probability Theory,

Vienna University of Technology, Austria

slide-2
SLIDE 2

data distribution per cell

Considering “scientific” data f, i.e.,

some data values f(p) (e.g., temperature, pressure values) measured/simulated wrt. a domain p (e.g., 2D/3D space, time, simulation input parameters)

If dimensionality of p > 3, then traditional visual analysis is hard Reducing the data dimensionality can help (e.g., computing statistical aggregates)

Higher-dimensional Scientific Data

Kehrer et al.: Brushing Moments in Interactive Visual Analysis

3D time-dependent multi-run simulation data

1

slide-3
SLIDE 3

Reducing the Data Dimensionality

Statistics: assess distributional

characteristics along an independent data dimension (e.g., time, spatial axes) Integrate aggregated statistics into visual analysis through attribute derivation

[from IPCC AR #4, 2007]

average temp. in ten years

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 2

slide-4
SLIDE 4

CLIMBER-2 model: Meltwater outburst of Lake Agassiz

3D atmosphere 250 time steps 240 runs (7 model parameters)  Compute local statistics wrt. multiple runs

Example: Multi-run Climate Simulation Data

timestep 80

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 3

slide-5
SLIDE 5

Moment-based Visual Analysis

Get big picture (data trends & outliers) Multitude of choices, e.g,

statistical moments (mean, std. deviation, skewness, kurtosis) traditional and 2 robust estimates compute relation (e.g., differences, ratio) change scale (e.g., data normalization, log. scaling, measure of “outlyingness”)

How to deal with this “management challenge”?

Kehrer et al.: Brushing Moments in Interactive Visual Analysis

4 ×3 ×2 ×3 = 72 possible configurations per axis

4

slide-6
SLIDE 6

Moment-based Visual Analysis

Kehrer et al.: Brushing Moments in Interactive Visual Analysis

quantile plot (focus+context)

multi-run data aggregated data

Iterative view transformations

alter axis/attribute configuration (construct a multitude of informative views) maintain mental model of views classification of moment-based views

Relate

multi-run data  aggregated data

5

slide-7
SLIDE 7

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 6

Visualizing Data Distributions

Sample quantile q(p) of a distribution {x1,…,xn}

at least n· p observations ≤ q(p) at least n·(1-p) observations ≥ q(p) Examlpes: median q2 = q(½), quartiles q1, q3

Quantile plot

shows all data items of a distribution assess data characteristics (normal dist., symmetrical, skewness, possible outliers, etc.)

6

quantile plot

q(p) p

(p ∈ [0, 1])

q3 q2 q1

normal distr.

q3 q2 q1

25% 25% 25% 25%

slide-8
SLIDE 8

Outlier influence traditional estimates Robust estimates of std. deviation

0.741· interquartile range (IQR) median absolute deviation MAD(x1,…,xn) = 1.483 · med1 ≤ i ≤ n (| xi – median |)

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 7

Robust Statistics

7

q1 q2 q3

interquartile range

mean median

slide-9
SLIDE 9

( )

∑ =

− =

n i n i MAD

x x MAD median x n skew

1 3 1 3

) ,..., ( 1

( )

∑ =

− =

n i i

s x x n skewness

1 3 3

1

Robust Statistics

Estimates of skewness

traditional median/MAD-based

  • ctile-based

Analogous estimates for kurtosis

Johannes Kehrer 8

1 7 4 1 7

2 e e e e e skewoct − − + =

  • neg. skewness

(left skewed)

  • pos. skewness

(right skewed)

mean median mean median

pos. kurtosis neg. kurtosis

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 8

slide-10
SLIDE 10

Iterative View Transformations

Change axis/attribute configuration of view

change order of moment robustify moment compute relation

(e.g., difference or ratio)

change scale

(e.g., normalize, z-standardization)

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 9

traditional med/MAD-based

  • ctile-based

Closer related to data tranformations

slide-11
SLIDE 11

change order of moment

 study relations

  • btw. moments

 investigate basic characteristics

  • f distributions

Basic View Setup: Opposing Different Moments

multi-run data aggregated data

quantile plot (focus+context) 1st vs. 2nd moment 3rd vs. 2nd moment 3rd vs. 4th moment

10

slide-12
SLIDE 12

Views: Opposing Different Moments

robustify moment

 assess influence

  • f outliers

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 11

traditional estimates

  • ctile-based

estimates

slide-13
SLIDE 13

Views: Traditional vs. Robust Estimates

robustify moment

 assess influence

  • f outliers

12

slide-14
SLIDE 14

Other View Transformations

compute relation

(e.g., difference or ratio)

change scale

(e.g., z-standardization, normalize to [0,1])

z-score (measure of

  • utlyingness)

distance to median

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 13

quantiles of

  • riginal data
slide-15
SLIDE 15

Other View Transformations

change scale  compare with theoretical distribution

Q-Q plot detrended Q-Q plot

x-axis subtracted from y-axis

Kehrer et al.: Brushing Moments in Interactive Visual Analysis 14

compute relation (e.g., difference)

slide-16
SLIDE 16

Analysis of Multi-run Climate Data

CLIMBER-2: Meltwater outburst of Lake Agassiz

(7 model parameters)

3D atmosphere 250 time steps 240 runs

15

slide-17
SLIDE 17

Kehrer et al.: Brushing Moments in Interactive Visual Analysis

Conclusion

Data aggregation enables complex analysis

relating higher-dimensional data  aggregated statistics analyst can work with both (original data, stat. properties)

Traditional and robust estimates of moments

(many opportunities also create management challenge)

Iterative view transformations

helps analyst to maintain a mental model of views matches the iterative nature of visual analysis

Classification of informative views

  • pposing different moments

traditional vs. robust estimates of same moment

16

slide-18
SLIDE 18

Kehrer et al.: Brushing Moments in Interactive Visual Analysis

Acknowledgements

Helmut Doleisch, Philipp Muigg www.SimVis.at Thomas Nocke, Michael Flechsig (PIK)

Datasets are courtesy of

VisGroup in Bergen, especially Armin Pobitzer, Stian Eikeland

17