Spatial Data Science - Chapter 2

Geovisualization
Luc Anselin
http://spatial.uchicago.edu
Copyright © 2016 by Luc Anselin, All Rights Reserved

• from EDA to ESDA
• from mapping to geovisualization
• mapping basics
• multivariate EDA primer

From EDA to ESDA

• Exploratory Data Analysis (EDA)
• reaction to modeling without looking at the data
• classic EDA book, Tukey (1977)
• Good (1983), Philosophy of Science
• “discover potentially explicable patterns”

• Data Visualization
• concept of a “view” (e.g., Buja et al 1996)
• a graphical representation and summary of the data
• many different views
• chart, table, graph, map

• Visual Explanations
• Tufte (1997) and later
• reasoning about evidence and design of graphics
• multivariate nature of analytic problems
• document sources (metadata)
• quantify and show cause and effect
• evaluate alternative explanations

• Visual Analytics
• Thomas et al (2005)
• the science of analytical reasoning facilitated by

interactive visual interfaces
• “detect the expected and discover the

unexpected”

• Exploratory Spatial Data Analysis (ESDA)
• EDA +
• describe spatial distributions
• dynamic statistical maps
• identify atypical spatial observations

• spatial outliers
• discover patterns of spatial dependence and

spatial heterogeneity
• spatial clusters, hot spots, cold spots
• spatial structural breaks

From Mapping to
Geovisualization

• What Is a Map
• “a collection of spatially defined
objects” (Monmonier)
• importance of depicting location
• importance of representing value

• How to Lie with Maps
• many design issues
• legends, colors, intervals
• projections
• human perception can be tricked
• political maps

http://xefer.com//2008/04/maps

• Geovisualization
• map + scientific visualization
• map as presentation vs map as part of the

analysis
• interactive mapping

• Maps and Knowledge Discovery
• exploration, synthesis, presentation, analysis
• visual popout
• abductive approach = pattern discovered along with

a hypothesis
• contrast with deductive or inductive
• interaction between data exploration and human

perception

• Geovisual Analytics
• leverages both geovisualization and visual
analytics
• interactive mapping
• animation
• linking and brushing

www.geovista.psu.edu
• Dynamic Graphics
• different views to represent the data
• focusing individual views
• linking multiple views
• arranging many views

• Linking and Brushing
• linking
• selection in one view (graph) is simultaneously

selected in all views
• brushing
• dynamically changing the selection updates all views

linked map and graph

Mapping Basics

• Choropleth Map
• not chloro!
• choros = region
• visualize a spatial distribution
• map counterpart of a histogram
• discrete approximation of the distribution
• all observations in the same value interval get the

same color

histogram and equal intervals choropleth map
• Choice of Intervals
• cut points
• equal interval, natural breaks (Jencks), manual
• statistical criteria
• equal share (quantile), standard deviational units

• Map Design Issues
• choice of colors
• perception of pattern
• red = hot, danger; blue = cool
• misleading role of area

• larger areas seem more important
• legends
• sequential
• diverging
• categorical
Statistical Maps

• Quantile Map
• data sorted from low to high
• equal number of observations in each interval
• examples
• quartile map (4 categories)
• quintile map (5 categories)
• possible issues with ties

quintile map (NYC % rental units)

• Box Map
• identifying outliers
• same principle as in box plot
• fence = median + 1.5 IQR or + 3 IQR
• IQR = inter quartile range, 25% to 75%
• six intervals
• same principle as quartile map
• outliers identified as a separate category

upper outliers in box plot and box map
(NYC median rent 2008)

lower outliers in box plot and box map

• Standard Deviational Map
• based on standardized data values
• mean = 0, standard deviation = 1
• intervals correspond to one standard deviation
• outliers are more than 2 standard deviations

from the mean

standard deviational map

• Cartogram
• areal unit proportional to variable of interest
• avoid misleading effect of area
• use transformed shapes
• circular cartogram
• contiguous cartogram

box map and circular cartogram
contiguous cartogram
area = number of votes in electoral college
source: Sarah Williams

• Conditional Maps
• cc maps, conditioned choropleth maps (Carr)
• special case of trellis graphs
• micromap matrix
• conditioning variables on the axes
• matrix of mini maps for the variable of interest

conditioned by the values on the axes

child malnutrition cc map conditioned on poverty index 
and per capita income (Nepal districts)

• Map Animation
• map movie
• highlight observations in increasing or

decreasing order
• one at a time
• cumulative
• visual impression of patterning/clustering

Multivariate EDA
Primer

• Objectives of Multivariate EDA
• represent multi-dimensional data in two
dimensions
• dimension reduction
• projection
• discover structure, interaction, patterns

• 3-D Scatter Plot
• points in a 3-D data cube
• two-dimensional analysis on side panels
• issues of perspective
• zooming, rotating
• brushing the 3-D data cube

selection in a 3D scatter plot

• Parallel Coordinate Plot (PCP)
• due to Inselberg (1984)
• variables
• one parallel line for each variable
• observations
• a line connecting points on the parallels
• the line is the counterpart of a point in the

multidimensional data cube

selected points in PCP

• Clusters in PCP
• lines that move closely together correspond to
points closely together in multidimensional
space
• = clusters
• visual cluster identification
• problems with large data sets
• remove clutter

brushing the PCP

• Scatter Plot Matrix
• matrix of bivariate scatter plots
• each variable once on x-axis and once on y-axis
• univariate description on diagonal
• focus on interaction effects

scatter plot matrix (Nepal districts)
scatter plot matrix with lowess smoother
brushing the scatter plot matrix

• Conditional Plots
• trellis display
• conditioning variables on the axes
• matrix of micro plots for subsets of

observations that match the axes conditions
• data intervals in two dimensions

scatter plot trellis graph
scatter of per capital income on no safe water
conditioned on poverty index and life expectancy
• Interpretation of Conditional Plots
• micro plots are similar
• no effect of conditioning variables
• micro plots are different
• conditioning variables interact with variable under

consideration
• effect of conditioning variables

Spatial Data Science - Chapter 2

Uploaded by

Copyright:

Available Formats

Spatial Data Science - Chapter 2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Spatial Data Science - Chapter 2

Uploaded by

Copyright:

Available Formats

Geovisualization

Copyright © 2016 by Luc Anselin, All Rights Reserved

Copyright © 2016 by Luc Anselin, All Rights Reserved

Copyright © 2016 by Luc Anselin, All Rights Reserved

• classic EDA book, Tukey (1977)

• Good (1983), Philosophy of Science

• “discover potentially explicable patterns”

Copyright © 2016 by Luc Anselin, All Rights Reserved

• a graphical representation and summary of the data

• many different views

• chart, table, graph, map

Copyright © 2016 by Luc Anselin, All Rights Reserved

• reasoning about evidence and design of graphics

• multivariate nature of analytic problems

• document sources (metadata)

• quantify and show cause and effect

• evaluate alternative explanations

Copyright © 2016 by Luc Anselin, All Rights Reserved

• the science of analytical reasoning facilitated by

• “detect the expected and discover the

Copyright © 2016 by Luc Anselin, All Rights Reserved

• identify atypical spatial observations

• discover patterns of spatial dependence and

Copyright © 2016 by Luc Anselin, All Rights Reserved

Copyright © 2016 by Luc Anselin, All Rights Reserved

• importance of depicting location

• importance of representing value

Copyright © 2016 by Luc Anselin, All Rights Reserved

• legends, colors, intervals

• human perception can be tricked

Copyright © 2016 by Luc Anselin, All Rights Reserved

Copyright © 2016 by Luc Anselin, All Rights Reserved

• map as presentation vs map as part of the

Copyright © 2016 by Luc Anselin, All Rights Reserved

• abductive approach = pattern discovered along with

• contrast with deductive or inductive

• interaction between data exploration and human

Copyright © 2016 by Luc Anselin, All Rights Reserved

• linking and brushing

Copyright © 2016 by Luc Anselin, All Rights Reserved

• focusing individual views

• linking multiple views

• arranging many views

Copyright © 2016 by Luc Anselin, All Rights Reserved

• selection in one view (graph) is simultaneously

• dynamically changing the selection updates all views

Copyright © 2016 by Luc Anselin, All Rights Reserved

Copyright © 2016 by Luc Anselin, All Rights Reserved

Copyright © 2016 by Luc Anselin, All Rights Reserved

• visualize a spatial distribution

• map counterpart of a histogram

• discrete approximation of the distribution

• all observations in the same value interval get the

Copyright © 2016 by Luc Anselin, All Rights Reserved

• equal interval, natural breaks (Jencks), manual

• equal share (quantile), standard deviational units

Copyright © 2016 by Luc Anselin, All Rights Reserved

• misleading role of area

Copyright © 2016 by Luc Anselin, All Rights Reserved

• equal number of observations in each interval