Spatial Patterns and Cartographic Danger

Question 1

3 basic spatial patterns

Answer 1

Random, uniform and clustered

Question 2

Cartographic Danger

Answer 2

Aggregate regional counts, often to preserve confidentiality
Counts lose some resolution
Different aggregations can lead to different results
Need to be critical with aggregated data maps
Question: what the underlying data is saying
E.g. county v ZIP Code v Census tract → changes what the message the map conveys
Can obscure heterogeneity at the larger spatial unit → smaller the spatial unit = more variation

Question 3

Modifiable areal unit problem (MAUP)

Answer 3

different aggregations can lead to diffrent results
Gerrymandering: Regional data and the modifiable areal unit problem
Congressional districts - measure number of people per district for representation in govt
Re-district to divvy up district to political advantage
Massaging teh data
Do not use congressional districts for health data as unreliable
Aiming to get good counts
Happens at state level

Question 4

Why is confidentiality important?

Answer 4

do not want to “out” someone if a disease with a lot of stigma + low prevalence → e.g. mask data or choose not to use household level data - choose your audience

Question 5

Areal Units: Census Geography

Answer 5

More Areal Units
States (US), counties, tracts (most commonly used by Dr Clennon)
Block groups
Blocks
We use areas because
Often summaries are only available for areas
Most common background demographic information in US is from the decennial census
Census data is free

Question 6

FIPS Code

Answer 6

Federal Information Processing Standard Codes
Hierarchical and can be broken apart or concatenated (combined)
Allows you to do table joins

Question 7

ZIP Code Areas

Answer 7

Designed to deliver mail, not tabulate the population
Commonly used because -
Can extract ZIP from billing address
Would have to geocode address to find tract, block, block-groups
Can approximate census tracts but not the same

Question 8

Measuring geographic distribution: Mean Centre

Answer 8

A way to mask data
Measures centre tendency
Average x + y values for input
Input locations need to be projected to accurately measure distances
Can compute locations or weight the locations by some numeric values
Mean centres can be compared to look at differences between population, case and incidence distribution

Question 9

Measuring Geographic Distribution: Median Centre

Answer 9

Measures central tendency but less influenced by outliers than the mean
Measures the distribution of location values around statistical mean
Uses an iterative algorithm to locate the point minimising the distance to all locations

Question 10

Measuring Geographic Distribution: Standard Deviation Ellipse

Answer 10

Shows the directional distribution - measures dispersion
Allows you to look at trends
1 standard deviation ellipse polygon will cover 68% of the points, 2SD = 95%, 3 SD = 99%