# Geographic Data Science - Lecture IV

## Today

• Mapping data
• MAUP
• Choropleths
• Definition
• Classes
• Cartograms
• Conditional maps
• Space-Time mapping

### Mapping Data with Data maps

• Abstraction from the purely geographical map
• Representing numerical values within a spatial context
• A geographical approach to statistical visualization

Key components

• “Container” geographies
• Choropleths: displaying data on maps (choropleths)

## MAUP

Modifiable Areal Unit Problem (Openshaw, 1984)

## MAUP

Scale and delineation mismatch between:

• Underlying process (e.g. individuals, firms, shops)
• Unit of measurement (e.g. neighborhoods, regions, etc.)

In some cases, it can seriously mislead analysis on aggregated data (e.g. Flint, MI!!!)

Always keep MAUP in mind when exploring aggregated data!!!

## Choropleths

Thematic map in which values of a variable are encoded using a color gradient of some sort
• Counterpart of the histogram
• Values are classified into specific colors: value –> bin
• Information loss as a trade off for simplicity

## Classification choices

• Colors –> in alignment with the goal of the map
• Bins –> How many?
• Algorithm:
• Unique values
• Equal interval
• Qua/Quintiles (equal count)
• Fisher-Jenks

## Color schemes

• Categories, non-ordered

[Source]

## Unique values

• Categorical data
• No gradient (reflect it with the color scheme!!!)
• Examples: Religion, country of origin…

## Equal interval

• Take the value span of the data to represent and split it equally
• Splitting happens based on the numerical value
• Gives more weight to outliers if the distribution is skewed

## Quantiles

• Regardless of numerical values, split the distribution keeping the same amount of values in each bin
• Splitting based on the rank of the value
• If distribution is skewed, it can put very different values in the same bin

## Other

• Fisher-Jenks
• Natural breaks
• Outlier maps: box maps, std. maps…

## Tips

Different classification schemes can produce widely different maps as a result of:

• The distribution of the values
• The inherent simplification that a choropleth implies

Best advice is to explore different ones and combine choropleths with other graphical devices like histograms or density plots

## Cartograms

Data maps where the variable is encoded, not by a color gradient, but by distorting the shape/size of the geographical objects”

• Useful in cases where the natural size/shape induces to wrong interpretation, or obscures the intended representation.
• If not done carefully, it can distort the message in unintended ways

[Source]

## Conditional maps

Split a dataset in buckets by conditioning on additional variables, then create a map for each bucket
• If no association, maps should look the same
• But, if the conditioning variables are somewhat related to the outcome we are mapping, the spatial distribution can vary substantially
• Exploration of multivariate relationships

## Space-Time mapping

• Bringing time into a spatial 2D context is “tricky” (it’s really 3D!)
• Traditionally –> sequence of time periods, 3D plots
• More recently –> animation and interactivity
[Source]

[Source]