Outlines 1. Terminology and User Tasks for time series 2. Limit
of Statistics Parameters 3. Examples by Tableau 4. Examples of
other tools 5. Design Principles and High Dimensionality Challenge
Objectives 1. Examination of a number of case studies 2. Learn from
some of the different visualization ideas that have been created 3.
Can you generalize these techniques into classes or
categories?
Slide 3
1. Time Series Data Fundamental chronological component to the
data set Random sample of 4000 graphics from 15 of world s
newspapers and magazines from 74-80 found that 75% of graphics
published were time series Tufte, Vol. 1
Slide 4
Taxonomy Time Series Discrete points vs. interval points Linear
time vs. cyclic time Ordinal time vs. continuous time Ordered time
vs. branching time vs. time with multiple perspectives Cross
Section Data (Multiple subjects/levels) Panel /TSCS (time-series
cross-sectional) Data
Slide 5
1. Terminology Data Sets Each data case is likely an event of
some kind e.g. sunspot activity, baseball games, medicines taken,
cities visited, stock prices, etc. One of the variables can be the
date and time of the event Other Data Attribute Prices Trade
volume
Slide 6
1. Meta Level Consider multiple stocks being examined Is each
stock a data case, or is a price on a particular day a case, with
the stock name as one of the other variables? Confusion between
data entity and data cases Answers to time series data for multiple
stocks or call 410K, Mutual fund profile Data entity Data Cases
Data attributes
Slide 7
1. Data Exploration Vs. Data Mining Data mining domain has
techniques for algorithmically examining time series data, looking
for patterns, etc. Good when objective is known a priori But what
if not? Which questions should I be asking? InfoVis and data
exploration better for that
Slide 8
1. Applications Autocorrelation analysis to examine Serial
Dependence Spectral Analysis to examine cyclic behavior which need
not be related to seasonality For example, sun spot activity varies
over 11 year cycles Other common examples include celestial
phenomena, weather patterns, neural activity, commodity prices, and
economic activity. Separation into components representing trend,
seasonality, slow and fast variation, and cyclical
irregularity
Slide 9
1. User tasks for time series What kinds of questions do people
ask about time series data? Examples When was something
greatest/least? Is there a pattern? Are two series similar? Do any
of the series match a pattern? Do some events have causal
relationships? Provide simpler, faster access to the series
Slide 10
1. Other Tasks Does data element exist at time t ? When does a
data element exist? How long does a data element exist? How often
does a data element occur? How fast are data elements changing? In
what order do data elements appear? Do data elements exist
together? Answer all these questions about stock price
Slide 11
1. Fundamental Tradeoff Is the visualization time-dependent,
i.e., changing over time (beyond just being interactive) Static
Shows history, multiple perspectives, allows comparison Dynamic
(animation) Gives feel for process & changes over time, has
more space to work with
Slide 12
Limit of Regress Analysis
Slide 13
Four Sets of Data IIIIIIIV xyxyxyxy 108.04109.14107.4686.58
86.9588.1486.7785.76 137.58138.741312.7487.71 98.8198.7797.1188.84
118.33119.26117.8188.47 149.96148.10148.8487.04
67.2466.1366.0885.25 44.2643.1045.391912.50
1210.84129.13128.1585.56 74.8277.2676.4287.91 55.6854.7455.7386.89
Anscombe's Quartet, American Statistician, 27 [February 1973],
17-21) comprises 4 data sets of 11 points each:
Slide 14
3.1 How about Linear Regress? Mean of the x values = 9.0 Mean
of the y values = 7.5 Equation of the least-squared regression line
is: y = 3 + 0.5x Sums of squared errors (about the mean) = 110.0
Regression sums of squared errors (variance accounted for by x) =
27.5 Residual sums of squared errors (about the regression line) =
13.75 Correlation coefficient = 0.82 Coefficient of determination =
0.67 http://astro.swarthmore.edu/astro121/anscombe.html
Slide 15
3.1. Four Data Sets
Slide 16
3. Basic Graphs of Time Series Present time data as a 2D line
graph with time on x- axis and some other variable on y-axis
Slide 17
3. Classic View
Slide 18
Fun example What If Everybody in Canada Flushed At Once?
Slide 19
3. A few examples by Tableau
Slide 20
3. Interactive Filter
Slide 21
Time Series segments for comparison
Slide 22
Comparison with new dimensions
Slide 23
3. Time series with distribution
Slide 24
Multiple time series with distribution
Slide 25
3. Monitor Multiple Views of Time series
Slide 26
4. Other tools, Example 1 Calendar visualization Present series
of events in context of calendar Task Put together complete story
Garner information for decision-making Notice trends Gain an
overview of the events to grasp the big picture
Slide 27
One Solution
Slide 28
3D View and Projections
Slide 29
Example 2 Personal histories Consider a chronological series of
events in someones life Present an overview of the events Examples
Medical history Educational background Criminal history Tasks Put
together complete story Gather information for decision-making
Notice trends Gain an overview of the events to grasp the big
picture
Slide 30
Life Line Project Visualize personal history in some
Domain
Slide 31
Medical display
Slide 32
Interaction Features Different colors for different event types
Line thickness can correspond to another variable Interaction:
Clicking on an event produces more details Certainly could also
incorporate some dynamic query capabilities
Slide 33
Benefit and Challenges Benefit Reduce chances of missing
information Facilitate spotting trends or anomalies Streamline
access to details Remain simple and tailorable to various
applications Challenges Scalability Can multiple records be
visualized in parallel (well)?
Slide 34
New works for Medical Image Work with query results Need to
align, rank, and filter Medical application: Look for temporal
coincidence of two events First pneuomonia and asthma attack
Medical professionals dont want to fool with zooming and
panning
Slide 35
Focus on alignment of events
Slide 36
Example 3 Understand patterns of presence/events over time
Focus: Peoples presence/movements in some space Situation: Workers
punch in and punch out of a factory Want to understand the presence
patterns over a calendar year Alternate: Power plant electricity
usage over a year
Slide 37
KW/Punch in/out times for workers
Slide 38
Approach Cluster analysis Find two most similar days, make into
one new composite Keep repeating until some preset number left or
some condition met How can this be visualized? Ideas?
Slide 39
Display
Slide 40
Characteristics Cluster Analysis Unique types of days
(individual or cluster) get their own color Contextually placed in
calendar and line graph for it is shown Stop clustering when a
threshold met or at a predetermined number of clusters Interactive
Click on day, see its graph Select a day, see similar ones
Add/remove clusters
Slide 41
Insight from the analysis Traditional office hours followed
Most employees present in late morning Fewer people are present on
summer Fridays Just a few people work holidays When the holidays
occurred School vacations occurred May 3-11, Oct 11-19, Dec 21-31
Many people take off day after holiday Many people leave at 4pm on
December 5
Slide 42
Example 4 Flow of changes across electronic documents history
flow is being used to visualize the evolutionary history of wiki*
pages on Wikipedia. http://researchweb.watson.ibm.com/history/
Slide 43
Techniques
Slide 44
What can be found? Understanding the frequency and timing of
vandalism for wiki sites. Analyzing the overall stability in size
and content for assessing the reliability of group-authored web
sites. Have a solid understanding of the relationship between
various factors. e.g., how does anonymity affect the likelihood of
vandalism? Are page sections that survive many edits more likely to
be high quality? The Visualizations above are suggestive, Need to
be verified through statistical analysis.
Slide 45
Example 5 Computer system logs Potentially huge amount of data
Tedious to examine the text Looking for unusual circumstances,
patterns, etc. MieLog System to help computer systems
administrators examine log files Interesting characteristics
Slide 46
System View
Slide 47
Another View
Slide 48
Interaction Features Tag area Click on tag shows only those
messages Time area Click on tiles to show those times Can put line
on histogram to filter on values above/below Outline area Can
filter based on message length Just highlight messages to show them
in text Message area Can filter on specific words
Slide 49
Example 6 Very large scale temporal log data Show more context
of what else was going on at that time Likely have to abstract some
then Allow several different levels of detail at once Allow
drill-down for details Domain: Computer systems management
Slide 50
LiveRAC Computer system management data Heavy interaction
Semantic zooming
Slide 51
5. Design Principles Show familiar visual representations
whenever possible Provide side-by-side comparisons of small
multiple views Spatial position is strongest visual cue Multiple
views are more effective when coordinated through explicit linking
Avoid abrupt visual change Follow Shneiderman s mantra User actions
should receive immediate visual feedback Assertion: Showing several
levels of detail simultaneously provides useful high information
density in context
Slide 52
Periodic Data Serial, periodic data Data with chronological
aspect, but repeats and follows a pattern over time How might one
visualize that? Using Spirals Standard x-y timeline or tabular
display is problematic for periodic data It has endpoints Use
spiral to help display data One loop corresponds to one period
Slide 53
Basic Spiral Graph One year per loop Same month on radial bars
Quantity represented by size of blob Is it as easy to see serial
data as periodic data?
Slide 54
Advanced Spiral Graph Spiral Same mapping as previous one
Different foods represented by different colors and drawn at
different heights Can you still see serial and periodic attributes?
As with all 3-D, requires navigation
Slide 55
Example 1, Geo-temporal data How about events in time and
place? Many applications of this problem Represent place by 2D
plane (or maybe 3D topography) Use 3 rd dimension to encode time
Object types: Entities (people or things) Locations (geospatial or
conceptual) Events (occurrences or discovered facts)
Slide 56
Overview Objective: visualize spatial interconnectedness of
information over time and geography with interactive 3-D view
Slide 57
Design Characteristics
Slide 58
Solution
Slide 59
March size of army directional altitude, longitude temperature,
date Napoleon s March
