Download Slides_VisualisingUncertainty

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Visualising Uncertainty
Dr Alasdair Rutherford
University of Stirling
What do we mean by
uncertainty?
.02
.01
.005
Density
.015
The simplest way to show
uncertainty is the
distribution of a variable
0
But showing all the data
doesn’t really summarise
20
40
60
age
80
100
Example: Box Plot
A box plot shows a number
of summary statistics about
spread that can help to
understand uncertainty
Use measures of spread as well as middle
Measures of the middle – mean, median, mode
Measures of spread – range, interquartile range,
standard deviation
Uncertainty in estimates – standard error and
confidence intervals
How do we communicate
uncertainty in the form of
probabilities?
Natural frequencies
appear to be superior to
percentages in improving
understanding of
screening tests
This is due to the
cognitive effort required
for interpreting
conditional probabilities
How do we communicate
uncertainty in predictions?
Compare two scenarios
Areas show
probability of
different outcomes
Different uncertainty on
different days
Show range of
predicted
temperatures, as
well as the average
Confidence
intervals around
estimates show
range of likely
values
Show both the
prediction and the
range of likely
possibilities
Fan-chart of
possible values
shows
uncertainty of
predictions –
and how this
changes over
time
Bank of England
refuses to
provide point
estimate for faroff predictions
so that the
focus is on the
uncertainty
What is the best way to visualise uncertainty?
Probabilities are
not intuitive –
communicate with
care!
Use multiple
formats; different Communicate
presentations suit both absolute and
different audiences relative
probabilities / risks
What is the best way to visualise uncertainty?
Well-presented
graphics with clear
explanations work
best
Use narratives,
images and
metaphors that gain
attention; but resist
arousing undue
emotion
All visualisations
have limitations,
showing only part of
the story – make it
clear what you are
and are not showing
Visualising Uncertainty
Communicating uncertainty is important,
whether you are reporting a simple average
or the estimates from statistical modelling
Visualisation can be a good way to
communicate these uncertainties – and there
are a wide range of techniques being
developed that can help
Visualising Uncertainty
Dr Alasdair Rutherford
University of Stirling