Visualizing Uncertainties
Kai Li
IST 719 “Information Visualization”
Nov. 11, 2013
The world is uncertain in nature
What is uncertainty?
or
Wh?t ?s unc?rt??nty?
Definitions of uncertainty
• A classification of statistical uncertainty:
–
–
–
–

Statistical variations or spread
Errors and differences
Minimum-maximum range values
Noisy or missing data (Pang, Wittenbrink, & Lodha, 1997)
(Pang et al., 1997)
Why should be care about
uncertainties?
Most, if not all stories are more complicated than it
looks
• The highest and lowest
kidney cancer death rates
happen in nearby counties,
which tend to be “rural,
mid-western, southern, and
western”. (Gelman, 2009)
• Is it because of any
geographical or
environmental factors?
Source: (Wainer, 2009)
We should, at least to some extent, expose the
complicity and uncertainties in the data
• Manuel Lima:
– Aspire for knowledge (Lima, 2009)

• Howard Wainer:
– Effective display of data must
• remind us that the data being displayed do contain some
uncertainty, and then
• characterize the size of that uncertainty as it pertains to the
inferences we have in mind, and in so doing
• help keep us from drawing incorrect conclusions through the
lack of a full appreciation of the precision of our knowledge.
(Wainer, 2009)
Examples of uncertainty visualization
• Traditional plots:
– Error bar
– Box plot and Violin plot
– Confidence/Prediction Intervals

• Visual cues that may be used:
–
–
–
–

Color
Blur
Glyph
Amplitude
Error Bar
• Error bars are a graphical
representation of the
variability of data and are
used on graphs to indicate
the error, or uncertainty in a
reported measurement.
– Pros:
• Effective way to present
errors and uncertainties
in the data
– Cons:
• Not appealing
Box Plot and Violin Plot
• Box Plot is a good way to
present groups of numerical
data through their quartiles
and outliers, thus to present
their variance and
uncertainty.

• Violin Plot is one of the
extensions to Box Plot, in
that it adds density of the
values to the x-axis in each
plot.
Confidence/Prediction Interval
• Confidence interval is a
range of values so defined
that there is a specified
probability that the value of
a parameter lies within it.
– A number of different models
to calculate confidence
interval.

• Prediction interval is the
range where you can expect
the next data point to
appear.
– A model is needed for
prediction.

(StackOverflow, n.d.)
Adding Extra Layers
Color

(Hengl, 2003)
Blur
• Pros:
– Blur is a preattentive visual
variable;
– It is also a perfect visual
metaphor for uncertain data.

• Cons:
– It’s hard to quantify blurry
areas.
(Kosara, 2001)
Adding glyph
• Adding glyph to vector field
to present uncertainty
information is common
especially for GIS information
visualization:
– Pros:
• Can be used to present
multi-facet uncertainty
information
• Will save more common
visual cues (color)
– Cons:
• Vector glyph can be visually
annoying

(Mahoney, 1999)
Amplitude modulation
• A. Cedilnik and P. Rheingans
used the density of
amplitude modulation in
annotation lines to mark the
uncertainty in each area.
(Cedilnik & Rheingans,
2000)
Questions
• 1. How can we integrate visualizing uncertainties into the
workflow of visualization design?
• 2. How to integrate uncertainty visualization to the bigger
graph to present meaningful information?
• 3. How can we evaluate the outcomes of uncertainty
visualization?
• 4. How can uncertainty visualization challenge the modernist
ways that stories are told using visualization?
– Is there a way to make visualization that
• Exposes the inaccuracy and discourse in the visualization per
se; or
• Deconstructs data/information in a meaningful way?
Reference
Andrej Cedilnik and Penny Rheingans (2000). Procedural Annotation of
Uncertain Information. Proceedings of IEEE Visualization '00, pp. 77-84.
Cedilnik, A., & Rheingans, P. (2000). Procedural annotation of uncertain
information. In Visualization 2000. Proceedings (pp. 77–84).
doi:10.1109/VISUAL.2000.885679
Gelman, A. (2004). Bayesian data analysis. Boca Raton, Fla.: Chapman &
Hall/CRC.
Hengl, T. (2003). Visualisation of uncertainty using the HSI colour model:
computations with colours. Retrieved November 10, 2013,
from http://www.academia.edu/1217951/Visualisation_of_uncertainty_u
sing_the_HSI_colour_model_computations_with_colours
Reference
Mahoney, D. P. (1999). The picture of uncertainty. Retrieved November 11,
2013, from http://www.cgw.com/Publications/CGW/1999/Volume-22Issue-11-November1999-/The-picture-of-uncertainty.aspx
Pang, A. T., Wittenbrink, C. M., & Lodha, S. K. (1997). Approaches to
uncertainty visualization. The Visual Computer, 13(8), 370–390.
doi:10.1007/s003710050111
StackOverflow. (n.d.). creating confidence area for normally distributed
scatterplot in ggplot2 and R. Retrieved November 10, 2013,
from http://stackoverflow.com/questions/7961865/creating-confidencearea-for-normally-distributed-scatterplot-in-ggplot2-and-r
Wainer, H. (2009). Picturing the Uncertainty world: How to understand,
communicate, and control uncertainty through graphical display.
Princeton: Princeton University Press.
Thank you!

Introduction to Visualizing Uncertainties

  • 1.
    Visualizing Uncertainties Kai Li IST719 “Information Visualization” Nov. 11, 2013
  • 2.
    The world isuncertain in nature
  • 3.
  • 4.
    Definitions of uncertainty •A classification of statistical uncertainty: – – – – Statistical variations or spread Errors and differences Minimum-maximum range values Noisy or missing data (Pang, Wittenbrink, & Lodha, 1997)
  • 5.
  • 6.
    Why should becare about uncertainties?
  • 7.
    Most, if notall stories are more complicated than it looks • The highest and lowest kidney cancer death rates happen in nearby counties, which tend to be “rural, mid-western, southern, and western”. (Gelman, 2009) • Is it because of any geographical or environmental factors?
  • 8.
  • 9.
    We should, atleast to some extent, expose the complicity and uncertainties in the data • Manuel Lima: – Aspire for knowledge (Lima, 2009) • Howard Wainer: – Effective display of data must • remind us that the data being displayed do contain some uncertainty, and then • characterize the size of that uncertainty as it pertains to the inferences we have in mind, and in so doing • help keep us from drawing incorrect conclusions through the lack of a full appreciation of the precision of our knowledge. (Wainer, 2009)
  • 10.
    Examples of uncertaintyvisualization • Traditional plots: – Error bar – Box plot and Violin plot – Confidence/Prediction Intervals • Visual cues that may be used: – – – – Color Blur Glyph Amplitude
  • 11.
    Error Bar • Errorbars are a graphical representation of the variability of data and are used on graphs to indicate the error, or uncertainty in a reported measurement. – Pros: • Effective way to present errors and uncertainties in the data – Cons: • Not appealing
  • 12.
    Box Plot andViolin Plot • Box Plot is a good way to present groups of numerical data through their quartiles and outliers, thus to present their variance and uncertainty. • Violin Plot is one of the extensions to Box Plot, in that it adds density of the values to the x-axis in each plot.
  • 13.
    Confidence/Prediction Interval • Confidenceinterval is a range of values so defined that there is a specified probability that the value of a parameter lies within it. – A number of different models to calculate confidence interval. • Prediction interval is the range where you can expect the next data point to appear. – A model is needed for prediction. (StackOverflow, n.d.)
  • 14.
  • 15.
  • 16.
    Blur • Pros: – Bluris a preattentive visual variable; – It is also a perfect visual metaphor for uncertain data. • Cons: – It’s hard to quantify blurry areas. (Kosara, 2001)
  • 17.
    Adding glyph • Addingglyph to vector field to present uncertainty information is common especially for GIS information visualization: – Pros: • Can be used to present multi-facet uncertainty information • Will save more common visual cues (color) – Cons: • Vector glyph can be visually annoying (Mahoney, 1999)
  • 18.
    Amplitude modulation • A.Cedilnik and P. Rheingans used the density of amplitude modulation in annotation lines to mark the uncertainty in each area. (Cedilnik & Rheingans, 2000)
  • 19.
    Questions • 1. Howcan we integrate visualizing uncertainties into the workflow of visualization design? • 2. How to integrate uncertainty visualization to the bigger graph to present meaningful information? • 3. How can we evaluate the outcomes of uncertainty visualization? • 4. How can uncertainty visualization challenge the modernist ways that stories are told using visualization? – Is there a way to make visualization that • Exposes the inaccuracy and discourse in the visualization per se; or • Deconstructs data/information in a meaningful way?
  • 20.
    Reference Andrej Cedilnik andPenny Rheingans (2000). Procedural Annotation of Uncertain Information. Proceedings of IEEE Visualization '00, pp. 77-84. Cedilnik, A., & Rheingans, P. (2000). Procedural annotation of uncertain information. In Visualization 2000. Proceedings (pp. 77–84). doi:10.1109/VISUAL.2000.885679 Gelman, A. (2004). Bayesian data analysis. Boca Raton, Fla.: Chapman & Hall/CRC. Hengl, T. (2003). Visualisation of uncertainty using the HSI colour model: computations with colours. Retrieved November 10, 2013, from http://www.academia.edu/1217951/Visualisation_of_uncertainty_u sing_the_HSI_colour_model_computations_with_colours
  • 21.
    Reference Mahoney, D. P.(1999). The picture of uncertainty. Retrieved November 11, 2013, from http://www.cgw.com/Publications/CGW/1999/Volume-22Issue-11-November1999-/The-picture-of-uncertainty.aspx Pang, A. T., Wittenbrink, C. M., & Lodha, S. K. (1997). Approaches to uncertainty visualization. The Visual Computer, 13(8), 370–390. doi:10.1007/s003710050111 StackOverflow. (n.d.). creating confidence area for normally distributed scatterplot in ggplot2 and R. Retrieved November 10, 2013, from http://stackoverflow.com/questions/7961865/creating-confidencearea-for-normally-distributed-scatterplot-in-ggplot2-and-r Wainer, H. (2009). Picturing the Uncertainty world: How to understand, communicate, and control uncertainty through graphical display. Princeton: Princeton University Press.
  • 22.