Data Visualization
Alternatives to maps
Andrew Irwin, a.irwin@dal.ca
Math & Stats, Dalhousie University
2021-03-19 (updated: 2021-03-06)
1 / 11
Plan
When is a map not appropriate for geospatial data?
Tile geometry
Heatmaps with dendrograms
Specialized heatmap for the USA
2 / 11
Why shouldn't I use a map?
Always think about the purpose
Is geography the most important feature?
Are you showing points or areas?
Will the area of a region complicate interpretation?
Is the spatial extent one dimensional? What could you do with the other dimension?
3 / 11
Summary
Maps are very effective for showing points where spatial location is the primary information
To show quantitative data varying along a transcect, a line or dot plot may be better
Colour shows larger/smaller, positive/negative, but does not show quantitative values well
Heatmaps (2 dimensional tiles of colours) can be a good alternative
4 / 11
COVID-19 cases & geom_tile
5 / 11
Same data with ggheatmap
6 / 11
Scale columns to have sum 1
7 / 11
Specialized maps of USA
library(statebins)load("../static/election.rda")m4 <- election %>% ggplot(aes(state = state, fill = pct_trump)) + geom_statebins() + theme_statebins() + labs(fill="Percent Trump")8 / 11
Specialized maps of USA
9 / 11
Further reading
Course notes
Healy Chapter 7
10 / 11
Task
- Task 16 as described in repository
11 / 11