All tracks

Spatial statistics

6 modules, 12.5 hours. Work through them in order: each module lists the ones it needs first.

1. s1 Bias in records

   • PostgreSQL with PostGIS
   • Browser
   • 1.5 hours

   Prerequisites: Spatial joins (g3)

   Skills: Datasets, basis of record, telemetry, recording effort; framing results honestly

   • How much of the jaguar sample is GPS-collar data?
   • Which datasets and years dominate a species' records?
   • What can records say about where an animal lives, and what can't they?

2. s2 Point patterns

   • PostgreSQL with PostGIS
   • Python
   • 2.5 hours

   Prerequisites: Bias in records (s1), Grids (g6)

   Skills: Quadrats, the nearest-neighbour index, kernel density, DBSCAN

   • Are devil records more clustered than chance?
   • Where is record density highest?

3. s3 Spatial weights

   • PostgreSQL with PostGIS
   • Python
   • 2 hours

   Prerequisites: Point patterns (s2)

   Skills: Distance-band, k-nearest and contiguity weights; row standardization; islands; ties

   • Who are a cell's neighbours?
   • What does row standardization do, and what happens to a cell with none?

4. s4 Global Moran's I

   • PostgreSQL with PostGIS
   • Python
   • 2 hours

   Prerequisites: Spatial weights (s3)

   Skills: Global Moran's I, its expectation, variance and z-score in SQL, matched by PySAL; permutation tests

   • Does record density cluster across Tasmania?
   • Do SQL and PySAL agree?
   • How does the answer change at 10 km?

5. s5 Hot and cold spots

   • PostgreSQL with PostGIS
   • Python
   • 2.5 hours

   Prerequisites: Global Moran's I (s4)

   Skills: Local Moran's I (LISA) and Getis-Ord Gi*; false discovery rate

   • Where are the hot and cold spots of recording effort?
   • Which survive a false discovery rate correction?

6. s6 Zonal statistics

   • PostgreSQL with PostGIS
   • Python
   • 2 hours

   Prerequisites: Hot and cold spots (s5), Rasters (g7)

   Skills: Zonal statistics on polygons and rasters; ratios normalized for effort

   • Do records over-represent some habitats or ecoregions, once effort is accounted for?