| CNTY_ | NAME | BIR74 | SID74 | NWBIR74 | BIR79 | SID79 | NWBIR79 |
|---|---|---|---|---|---|---|---|
| 1825 | Ashe | 1091 | 1 | 10 | 1364 | 0 | 19 |
| 1827 | Alleghany | 487 | 0 | 10 | 542 | 3 | 12 |
| 1828 | Surry | 3188 | 5 | 208 | 3616 | 6 | 260 |
| 1831 | Currituck | 508 | 1 | 123 | 830 | 2 | 145 |
| 1832 | Northampton | 1421 | 9 | 1066 | 1606 | 3 | 1197 |
| 1833 | Hertford | 1452 | 7 | 954 | 1838 | 5 | 1237 |
| 1834 | Camden | 286 | 0 | 115 | 350 | 2 | 139 |
10/16/2019, Spatial Data Science Conference, NY; https://edzer.github.io/sdsc19/ (repo)
Data science: the table view
… with coordinates
| CNTY_ | NAME | BIR74 | SID74 | longitude | latitude |
|---|---|---|---|---|---|
| 1825 | Ashe | 1091 | 1 | -81.49826 | 36.43140 |
| 1827 | Alleghany | 487 | 0 | -81.12515 | 36.49101 |
| 1828 | Surry | 3188 | 5 | -80.68575 | 36.41252 |
| 1831 | Currituck | 508 | 1 | -76.02750 | 36.40728 |
| 1832 | Northampton | 1421 | 9 | -77.41056 | 36.42228 |
| 1833 | Hertford | 1452 | 7 | -76.99478 | 36.36145 |
| 1834 | Camden | 286 | 0 | -76.23435 | 36.40120 |
… with POINT geometries
| CNTY_ | NAME | BIR74 | SID74 | geometry |
|---|---|---|---|---|
| 1825 | Ashe | 1091 | 1 | POINT (-81.49826 36.4314) |
| 1827 | Alleghany | 487 | 0 | POINT (-81.12515 36.49101) |
| 1828 | Surry | 3188 | 5 | POINT (-80.68575 36.41252) |
| 1831 | Currituck | 508 | 1 | POINT (-76.0275 36.40728) |
| 1832 | Northampton | 1421 | 9 | POINT (-77.41056 36.42228) |
| 1833 | Hertford | 1452 | 7 | POINT (-76.99478 36.36145) |
| 1834 | Camden | 286 | 0 | POINT (-76.23435 36.4012) |
… with POLYGON geometries
| CNTY_ | NAME | BIR74 | SID74 | geometry |
|---|---|---|---|---|
| 1825 | Ashe | 1091 | 1 | MULTIPOLYGON (((-81.47276 36.23436, -81.54084 3… |
| 1827 | Alleghany | 487 | 0 | MULTIPOLYGON (((-81.23989 36.36536, -81.24069 3… |
| 1828 | Surry | 3188 | 5 | MULTIPOLYGON (((-80.45634 36.24256, -80.47639 3… |
| 1831 | Currituck | 508 | 1 | MULTIPOLYGON (((-76.00897 36.3196, -76.01735 36… |
| 1832 | Northampton | 1421 | 9 | MULTIPOLYGON (((-77.21767 36.24098, -77.23461 3… |
| 1833 | Hertford | 1452 | 7 | MULTIPOLYGON (((-76.74506 36.23392, -76.98069 3… |
| 1834 | Camden | 286 | 0 | MULTIPOLYGON (((-76.00897 36.3196, -75.95718 36… |
the "shapefile" view
## Coordinate Reference System: ## EPSG: 4267 ## proj4string: "+proj=longlat +datum=NAD27 +no_defs"
Coordinate reference systems
- are the "measurement units" of spatial coordinates
- relate a location to a particular reference ellipsoid ("datum")
- may describe a 2-D projection
pts = st_centroid(st_geometry(nc))
## Warning in st_centroid.sfc(st_geometry(nc)): st_centroid does not give ## correct centroids for longitude/latitude data
st_crs(pts)
## Coordinate Reference System: ## EPSG: 4267 ## proj4string: "+proj=longlat +datum=NAD27 +no_defs"
pts[[1]]
## POINT (-81.49826 36.4314)
st_transform(pts[1], "+proj=longlat +datum=WGS84")[[1]]
## POINT (-81.49809 36.43152)
Joining tables, spatially
- if
longitudeandlatitudeare in columns (fields), then regularjoinwill work - if you have
POINTgeometries in a column, usest_equals(orst_within_distance) - if you have other geometries, there are a lot of options:
st_intersects,st_disjointst_covers,st_covered_by,st_touches,st_within- …
st_relatewith pattern
Machine learning with coordinates as features
Potential issues:
- you are trying to explain variability by where, instead of (or in addition to, or as an alternative to) what is going on there (feature variables); this may be competitive, but might be so for the wrong reason (extrapolation?)
- in case you have clustered training data and use naive (random) cross validation, location may pretty well point out the cluster, and result in overly optimistic performance measures
- as with regression modelling strategies, check that the models are not sensitive to translations and rotations of coordinates (in \(R^2\), or \(S^2\), or both?)