[PDF] - ggmap Available on CRAN! Recent Advances in Spatial Visualization PDF Document

SLIDE 1

ggmap

Recent Advances in Spatial Visualization

Department of Statistical Science

David J. Kahle, Ph.D.

Available on CRAN!

@

SLIDE 2

components of this talk

1. ggplot2
2. ggmap basics
3. Some modeling
4. Utility functions
5. Future directions

and the layered grammar of graphics ggmap = ggplot2 + online mapping sources

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

Visualizing spatial interpolation Updates to geocode

SLIDE 3

ggplot2

and the layered grammar of graphics

SLIDE 4 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the diamonds dataset

head(diamonds)

## carat cut color clarity depth table price x y z ## 1 0.23 Ideal E SI2 61.5 55 326 3.95 3.98 2.43 ## 2 0.21 Premium E SI1 59.8 61 326 3.89 3.84 2.31 ## 3 0.23 Good E VS1 56.9 65 327 4.05 4.07 2.31 ## 4 0.29 Premium I VS2 62.4 58 334 4.20 4.23 2.63 ## 5 0.31 Good J SI2 63.3 58 335 4.34 4.35 2.75 ## 6 0.24 Very Good J VVS2 62.8 57 336 3.94 3.96 2.48

SLIDE 5 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the diamonds dataset

Clarity Flawless Internally Flawless Very Very Slightly Included Very Slightly Included Slightly Included Included Grade FL IF VVS1 VVS2 VS1 VS2 SI1 SI2 I1 I2 I3

Color Clarity

Source : http://www.affordablediamondsonline.com/diamondshapes.jpg, http://beaufortsjeweler.com/images/diamond_color_chart.gif, http://www.am-diamonds.com/images/anatomy_diamond.gif

z table width x carat y price also depth = z / diameter table = table width / x

Everything else Cut Ideal > Premium > Very Good > Good > Fair

SLIDE 6 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the diamonds dataset

head(diamonds)

## carat cut color clarity depth table price x y z ## 1 0.23 Ideal E SI2 61.5 55 326 3.95 3.98 2.43 ## 2 0.21 Premium E SI1 59.8 61 326 3.89 3.84 2.31 ## 3 0.23 Good E VS1 56.9 65 327 4.05 4.07 2.31 ## 4 0.29 Premium I VS2 62.4 58 334 4.20 4.23 2.63 ## 5 0.31 Good J SI2 63.3 58 335 4.34 4.35 2.75 ## 6 0.24 Very Good J VVS2 62.8 57 336 3.94 3.96 2.48

library(dplyr) set.seed(1) d <- diamonds %>% sample_n(100)

the diamonds dataset is pretty big, so let’s thin it down:

SLIDE 7 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

plot(d$carat, d$price)

●
5000

10000 15000 0.5 1.0 1.5 2.0

d$carat d$price

●
0.5

1.0 1.5 2.0 5000 10000 15000 d$carat d$price

the ggplot2 framework

SLIDE 8

5000

10000 15000 0.5 1.0 1.5 2.0

d$carat d$price

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

plot(d$carat, d$price)

the ggplot2 framework

library(ggplot2) theme_set(theme_bw()) qplot(d$carat, d$price)

SLIDE 9

5000

10000 15000 0.5 1.0 1.5 2.0

d$carat d$price

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

plot(d$carat, d$price)

the ggplot2 framework

library(ggplot2) theme_set(theme_bw()) qplot(d$carat, d$price)

5000

10000 15000 0.5 1.0 1.5 2.0

carat price

qplot(carat, price, data = d)

SLIDE 10

qplot(carat, price, data = d)

5000

10000 15000 0.5 1.0 1.5 2.0

d$carat d$price

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the ggplot2 framework

5000

10000 15000 0.5 1.0 1.5 2.0

carat price

qplot(carat, price, data = d, geom = "point" )

SLIDE 11 5000 10000 15000 0.5 1.0 1.5 2.0

carat price

qplot(carat, price, data = d)

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the ggplot2 framework

qplot(carat, price, data = d, geom = "point" ) qplot(carat, price, data = d, geom = "smooth" )

SLIDE 12 5000 10000 0.5 1.0 1.5

carat price

qplot(carat, price, data = d)

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the ggplot2 framework

qplot(carat, price, data = d, geom = "point" ) qplot(carat, price, data = d, geom = "smooth" ) qplot(carat, price, data = d, geom = "density2d" )

SLIDE 13

5000

10000 15000 0.5 1.0 1.5 2.0

carat price

qplot(carat, price, data = d)

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the ggplot2 framework

qplot(carat, price, data = d, geom = "point" ) qplot(carat, price, data = d, geom = "smooth" ) qplot(carat, price, data = d, geom = "density2d" ) qplot(carat, price, data = d, geom = c("density2d", "point") )

SLIDE 14 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

abline bar histogram bin2d blank boxplot crossbar dotplot errorbar freqpoly hex hline linerange map density contour density2d path step point jitter

pointrange

polygon quantile raster rect ribbon rug segment smooth text tile violin vline

See docs.ggplot2.org/current/ for details

list of geoms

The geom you use depends on your situation!

SLIDE 15 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

abline bar histogram bin2d blank boxplot crossbar dotplot errorbar freqpoly hex hline linerange map density contour density2d path step point jitter

pointrange

polygon quantile raster rect ribbon rug segment smooth text tile violin vline

See docs.ggplot2.org/current/ for details

1d discrete geom

list of geoms

SLIDE 16 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

abline bar histogram bin2d blank boxplot crossbar dotplot errorbar freqpoly hex hline linerange map density contour density2d path step point jitter

pointrange

polygon quantile raster rect ribbon rug segment smooth text tile violin vline

See docs.ggplot2.org/current/ for details

1d continuous geoms

list of geoms

SLIDE 17 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

abline bar histogram bin2d blank boxplot crossbar dotplot errorbar freqpoly hex hline linerange map density contour density2d path step point jitter

pointrange

polygon quantile raster rect ribbon rug segment smooth text tile violin vline

See docs.ggplot2.org/current/ for details

discrete-continuous geoms

list of geoms

SLIDE 18 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

abline bar histogram bin2d blank boxplot crossbar dotplot errorbar freqpoly hex hline linerange map density contour density2d path step point jitter

pointrange

polygon quantile raster rect ribbon rug segment smooth text tile violin vline

See docs.ggplot2.org/current/ for details

continuous-continuous geoms

list of geoms

SLIDE 19

qplot(carat, price, data = d)

5000

10000 15000 0.5 1.0 1.5 2.0

d$carat d$price

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetics

5000

10000 15000 0.5 1.0 1.5 2.0

carat price

x y

SLIDE 20

5000

10000 15000 0.5 1.0 1.5 2.0

carat price

cut

Fair

Good Very Good Premium Ideal

qplot(carat, price, data = d)

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetics x y

qplot(carat, price, data = d, color = cut )

x y color

SLIDE 21

●
5000

10000 15000 0.5 1.0 1.5 2.0

carat price

cut

Fair

Good Very Good Premium Ideal

qplot(carat, price, data = d)

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetics x y

qplot(carat, price, data = d, color = cut )

x y color

qplot(carat, price, data = d, size = cut )

x y size

SLIDE 22 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetic mappings and the grammar of graphics

Variables in dataset

cut color clarity price

z measurement y measurement x measurement table depth x axis y axis z axis? size shape color fill alpha

pacity

Visual elements of graphic (aesthetics)

carat

SLIDE 23 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetic mappings and the grammar of graphics

Variables in dataset

cut color clarity price

z measurement y measurement x measurement table depth x axis y axis z axis? size shape color fill alpha

pacity

Visual elements of graphic (aesthetics)

carat

SLIDE 24

●
5000

10000 15000 0.5 1.0 1.5 2.0

carat price

cut

Fair

Good Very Good Premium Ideal

qplot(carat, price, data = d)

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetics

qplot(carat, price, data = d, color = cut ) qplot(carat, price, data = d, size = cut )

SLIDE 25 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetic mappings and the grammar of graphics

adj alpha angle bg cex col color colour fg fill group hjust label linetype lower lty lwd max middle min

rder

pch radius

sample

shape size srt upper vjust weight width x xend xmax xmin xintercept y yend ymax ymin yintercept z

different aesthetics apply to different geoms

SLIDE 26 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetic scales A scale is a function that maps levels of a variable to aesthetic values

cut color price

z measurement y measurement x measurement table depth x axis y axis z axis? size shape color fill alpha

pacity

carat

cut

Fair

Good Very Good Premium Ideal

Fair Good … … scale

SLIDE 27 5000 10000 15000 0.5 1.0 1.5 2.0

carat price

cut Fair Good Very Good Premium Ideal

qplot(carat, price, data = d)

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

aesthetics

qplot(carat, price, data = d, color = cut ) qplot(carat, price, data = d, size = cut, alpha = I(.4) )

setting aesthetics

for when you don’t want to map them to a variable

SLIDE 28 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

Discrete Continuous Color/Fill Size Shape Alpha

Rainbow of colors Color gradient Discrete size steps Different shape for points

Aesthetic effect by variable type

Continuous size gradient

Bubble Charts

Discrete transparency steps Continuous transparency gradient

aesthetic scales

SLIDE 29 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

layered ggplot specifications Both of the following make the same graphic

qplot(carat, price, data = d) ggplot() + layer( data = d, mapping = aes(x = carat, y = price), geom = "point" )

lays down a background and axes lays down a layer with only the geom

SLIDE 30 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

layered ggplot specifications

5000 10000 15000 0.5 1.0 1.5 2.0 carat price

●
ggplot()

layer(...) + =

●
5000

10000 15000 0.5 1.0 1.5 2.0 carat price

qplot(...)

SLIDE 31 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

layered ggplot specifications

= ggplot() layer(...) + qplot(...)

5000 10000 15000 0.5 1.0 1.5 2.0 carat price

●
●
●
5000

10000 15000 0.5 1.0 1.5 2.0 carat price

+ layer(...)

ggplot2 demands the aesthetic scales stay consistent across layers

SLIDE 32 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

layered ggplot specifications Both of the following make the same graphic

qplot(carat, price, data = d) ggplot() + layer( data = d, mapping = aes(x = carat, y = price), geom = "point" )

lays down a background and axes lays down a layer with only the geom

SLIDE 33

ggplot() + layer( data = d, mapping = aes(x = carat, y = price), geom = "point" ) + layer( data = d, mapping = aes(x = carat, y = price), geom = "density2d" )

David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

layered ggplot specifications Both of the following make the same graphic

qplot(carat, price, data = d)

SLIDE 34 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

layered ggplot specifications Both of the following make the same graphic

qplot(carat, price, data = d) ggplot() + geom_point(aes(carat, price), data = d) + geom_density2d(aes(carat, price), data = d)

SLIDE 35

ggmap

SLIDE 36 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

ggmap – the key ideas

ggplot() is replaced by ggmap() with a get_map() inside
the x aesthetic is always longitude, the y latitude
the projection is always mercator
the entire range of ggplot2 functionality works
qplot() is replaced by qmplot()

SLIDE 37 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

From the meuse dataset documentation (in sp)

This data set gives locations and topsoil heavy metal concentrations, along with a number of soil and l a n d s c a p e v a r i a b l e s a t t h e

bservation locations, collected in a

flood plain of the river Meuse, near the village of Stein (NL). Heavy metal concentrations are from composite samples of an area of approximately 15 m x 15 m.

Field data were collected by Ruud van Rijn and Mathieu Rikken; compiled for R by Edzer Pebesma; description extended by David Rossiter

the meuse dataset

SLIDE 38 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

head(m) ## lon lat cadmium copper lead zinc ## 1 5.758536 50.99156 11.7 85 299 1022 ## 2 5.757863 50.99109 8.6 81 277 1141 ## 3 5.759855 50.99089 6.5 68 199 640 ## 4 5.761746 50.99041 2.6 81 116 257 ## 5 5.761863 50.98903 2.8 48 117 269 ## 6 5.763040 50.98839 3.0 61 137 281

+ other variables the meuse dataset

SLIDE 39 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

when meuse is a spatialPointsDataFrame –

plot(meuse)

library(sp) library(rgdal)

!

data(meuse) coordinates(meuse) <- c("x", "y") proj4string(meuse) <- CRS("+init=epsg:28992") meuse <- spTransform(meuse, CRS("+proj=longlat +datum=WGS84"))

the meuse dataset

SLIDE 40 ## Map from URL : http://tile.stamen.com/toner-lite/15/16904/10970.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16905/10970.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16906/10970.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16907/10970.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16908/10970.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16904/10971.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16905/10971.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16906/10971.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16907/10971.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16908/10971.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16904/10972.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16905/10972.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16906/10972.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16907/10972.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16908/10972.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16904/10973.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16905/10973.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16906/10973.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16907/10973.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16908/10973.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16904/10974.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16905/10974.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16906/10974.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16907/10974.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16908/10974.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16904/10975.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16905/10975.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16906/10975.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16907/10975.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16908/10975.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16904/10976.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16905/10976.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16906/10976.png ## Map from URL : http://tile.stamen.com/toner-lite/15/16907/10976.png David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

qmplot when meuse is a data.frame –

qmplot(lon, lat, data = m)

●
●
## Using zoom = 15...

Map from URL : http://tile.stamen.com/toner-lite/15/16904/10970.png

# after the code from the last slide m <- data.frame(meuse@coords, meuse@data) names(m)[1:2] <- c("lon", "lat")

SLIDE 41 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

when meuse is a data.frame –

qmplot(lon, lat, data = m) ## Using zoom = 15...

●
●
qmplot(lon, lat, data = m,

zoom = 14 )

qmplot

SLIDE 42 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

qmplot(lon, lat, data = m, size = zinc, zoom = 14, source = "google", maptype = "satellite", alpha = I(.75), color = I("green"), legend = "topleft", darken = .2 ) + scale_size("Zinc (ppm)")

qmplot

SLIDE 43

Zinc (ppm) 500 1000 1500

SLIDE 44 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

qmplot(lon, lat, data = m, size = zinc, zoom = 14, source = "google", maptype = "satellite", alpha = I(.75), color = I("green"), legend = "topleft", darken = .2 ) + scale_size("Zinc (ppm)") + geom_density2d(color = "lightblue", size = .25)

Not helpful, but illustrative… qmplot with an additional layer

SLIDE 45

Zinc (ppm) 500 1000 1500

SLIDE 46 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

the base layer formerly the ggmap()

Long form

ggmap(get_map(...), ...)

Short form

qmap(...)

"ggmap" object

raster with meta data

get_map() ggmap()

Quartz 2 [*]

qmap()

…

Longest form

ggmap(get_googlemap(...), ...) ggmap(get_openstreetmap(...), ...) ggmap(get_stamenmap(...), ...) ggmap(get_cloudmademap(...), ...) ggmap(get_navermap(...), ...)

SLIDE 47 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

ggmap and the base layer the Stamen Design maptypes

SLIDE 48 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

list of geoms useful in ggmap

abline bar histogram bin2d blank boxplot crossbar dotplot errorbar freqpoly hex hline linerange map density contour density2d path step point jitter

pointrange

polygon quantile raster rect ribbon rug segment smooth text tile violin vline

See docs.ggplot2.org/current/ for details

SLIDE 49

some modeling

SLIDE 50 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

meuse.grid (in gstat) is a grid over the same space

plot(meuse.grid)

the meuse.grid dataset

library(gstat) data(meuse.grid) coordinates(meuse.grid) <- c("x", "y") proj4string(meuse.grid) <- CRS(“+init=epsg:28992") meuse.grid <- spTransform(meuse.grid, CRS("+proj=longlat +datum=WGS84"))

SLIDE 51

● ●
●
●
Zinc (ppm)
500

1000 1500 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

qmplot(lon, lat, data = mg, zoom = 14, legend = "topleft" ) + geom_point( aes(size = zinc), shape = I(15), data = m, color = "red" ) + scale_size("Zinc (ppm)")

What are the zinc levels at

ther locations?

mg <- data.frame(meuse.grid@coords, meuse.grid@data) names(mg)[1:2] <- c("lon", "lat")

SLIDE 52 500 1000 1500 Predicted Zinc (ppm) David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

qmplot(lon, lat, data = mg, shape = I(15), color = idw, zoom = 14, legend = "topleft", alpha = I(.75), darken = .4 ) + scale

inverse distance weighted interpolation

idw_mod <- idw(log(zinc) ~ 1, meuse, meuse.grid, idp = 2.5 ) mg$idw <- exp( idw_mod@data$var1.pred )

scale <- scale_color_gradient("Predicted\nZinc (ppm)", low = "green", high = "red", lim = c(100, 1850) )

SLIDE 53 500 1000 1500 Predicted Zinc (ppm) David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

linear regression

lin <- krige(log(zinc) ~ 1, meuse, meuse.grid ) mg$lin <- exp( lin@data$var1.pred ) qmplot(lon, lat, data = mg, shape = I(15), color = lin, zoom = 14, legend = "topleft", alpha = I(.75), darken = .4 ) + scale

SLIDE 54 500 1000 1500 Predicted Zinc (ppm) David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

trend surface analysis

tsa <- krige(log(zinc) ~ 1, meuse, meuse.grid, degree = 2 ) mg$tsa <- exp( tsa@data$var1.pred ) qmplot(lon, lat, data = mg, shape = I(15), color = tsa, zoom = 14, legend = "topleft", alpha = I(.75), darken = .4 ) + scale

SLIDE 55 500 1000 1500 Predicted Zinc (ppm) David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

rdinary kriging

vgram <- variogram(log(zinc) ~ 1, meuse) vgramFit <- fit.variogram(vgram, vgm(1, "Exp", .2, .1) )

rdKrige <- krige(log(zinc) ~ 1,

meuse, meuse.grid, vgramFit ) mg$ordKrige <- exp(

rdKrige@data$var1.pred

)

qmplot(lon, lat, data = mg, shape = I(15), color = ordKrige, zoom = 14, legend = "topleft", alpha = I(.75), darken = .4 ) + scale

SLIDE 56 500 1000 1500 Predicted Zinc (ppm) David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

universal kriging

vgram <- variogram(log(zinc) ~ 1, meuse) vgramFit <- fit.variogram(vgram, vgm(1, "Exp", .2, .1) ) univKrige <- krige(log(zinc) ~ sqrt(dist), meuse, meuse.grid, vgramFit ) mg$univKrige <- exp( univKrige@data$var1.pred )

qmplot(lon, lat, data = mg, shape = I(15), color = univKrige, zoom = 14, legend = "topleft", alpha = I(.75), darken = .4 ) + scale

SLIDE 57 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

faceting

IDW TSA

Ord. Krig.
Lin. Reg.
Uni. Krig.

500 1000 1500 Predicted Zinc (ppm)

library(dplyr); library(tidyr) mg2 <- mg %>% select(lon, lat, idw, tsa, ordKrige, lin, univKrige) names(mg2)[3:7] <- c("IDW", "TSA", "Ord. Krig.", "Lin. Reg.", "Uni. Krig.")

!

mg2 <- mg2 %>% gather("Method", "Prediction", 3:7)

qmplot(lon, lat, data = mg2, shape = I(15), color = Prediction, zoom = 14, alpha = I(.75), darken = .4 ) + scale + facet_wrap(~ Method)

SLIDE 58

utility functions

SLIDE 59 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

1. geocode()
2. revgeocode()
3. mapdist()
4. route()

SLIDE 60 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

geocode() Address Map coordinates Geocoding:

geocode("one bear place, waco, texas") ## lon lat ## 1 -97.16013 31.41688

Information from URL : http://www.datasciencetoolkit.org/maps/api/geocode/json?address=one+bear+place,+waco,+texas&sensor=false

geocode( c("baylor university", "the white house"), source = "google" ) ## lon lat ## 1 -97.11844 31.54822 ## 2 -77.03653 38.89768

SLIDE 61 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

geocode()

big12

## university address wins ## 1 baylor One Bear Place, Waco, TX 76798 8 ## 2 iowa state Ames, IA 50011 0 ## 3 kansas 1450 Jayhawk Boulevard, Lawrence, KS 66045 1 ## 4 kansas state Manhattan, KS 66506 7 ## 5 oklahoma 660 Parrington Oval, Norman, OK 73019 5 ## 6 oklahoma state Stillwater, OK 74074 4 ## 7 texas Austin, TX 78712 5 ## 8 texas christian 2800 South University Drive, Fort Worth, TX 76129 8 ## 9 texas tech 2500 Broadway, Lubbock, TX 79409 2 ## 10 west virginia Morgantown, WV 26506 5

SLIDE 62 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

geocode()

geocode(address, data = big12)

verwriting dataset big12.

big12

## university address wins lon lat ## 1 baylor One Bear Place, Waco, TX 76798 8 -97.16013 31.41688 ## 2 iowa state Ames, IA 50011 0 -93.63645 42.02403 ## 3 kansas 1450 Jayhawk Boulevard, Lawrence, KS 66045 1 -95.24935 38.95830 ## 4 kansas state Manhattan, KS 66506 7 -96.58388 39.19603 ## 5 oklahoma 660 Parrington Oval, Norman, OK 73019 5 -97.44528 35.21036 ## 6 oklahoma state Stillwater, OK 74074 4 -97.06906 36.10152 ## 7 texas Austin, TX 78712 5 -97.73100 30.28217 ## 8 texas christian 2800 South University Drive, Fort Worth, TX 76129 8 -97.36044 32.71171 ## 9 texas tech 2500 Broadway, Lubbock, TX 79409 2 -101.87063 33.58444 ## 10 west virginia Morgantown, WV 26506 5 -79.96267 39.64528

qmplot()

SLIDE 63 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

route()

legs_df <- route( 'marrs mclean science, baylor university', '220 south 3rd street, waco, tx 76701', alternatives = TRUE)

## m km miles seconds minutes hours startLon startLat endLon endLat leg route ## 1 64 0.064 0.039 6 0.10 0.0016 -97.11969 31.54889 -97.12016 31.54849 1 A ## 2 1183 1.183 0.735 184 3.06 0.0511 -97.12016 31.54849 -97.12902 31.55596 2 A ## 3 129 0.129 0.080 19 0.31 0.0052 -97.12902 31.55596 -97.12805 31.55678 3 A ## 4 35 0.035 0.021 21 0.35 0.0058 -97.12805 31.55678 -97.12831 31.55700 4 A ## 5 70 0.070 0.043 6 0.10 0.0016 -97.11969 31.54889 -97.11916 31.54933 1 B ## 6 208 0.208 0.129 51 0.85 0.0141 -97.11916 31.54933 -97.12072 31.55065 2 B ## 7 224 0.224 0.139 25 0.41 0.0069 -97.12072 31.55065 -97.11915 31.55214 3 B ## 8 1003 1.003 0.623 126 2.10 0.0350 -97.11915 31.55214 -97.12612 31.55845 4 B ## 9 260 0.260 0.161 42 0.70 0.0116 -97.12612 31.55845 -97.12805 31.55678 5 B ## 10 35 0.035 0.021 14 0.23 0.0038 -97.12805 31.55678 -97.12831 31.55700 6 B

SLIDE 64 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

route()

qmap('424 clay avenue, waco, tx', zoom = 15, maptype = 'hybrid', base_layer = ggplot( aes(x = startLon, y = startLat), data = legs_df) ) + geom_leg( aes( x = startLon, y = startLat, xend = endLon, yend = endLat, colour = route ), alpha = 3/4, size = 2, data = legs_df ) + facet_wrap(~ route, ncol = 2)

SLIDE 65 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

route()

A B route A B

SLIDE 66

future directions

SLIDE 67 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

1. boost geocode
2. better modeling integration
3. new graphical packages

mutate_geocode(), more sources and failsafes something like geom_krige()

SLIDE 68 David J. Kahle, Ph.D. 45th Symposium on the Interface (2015) : Data Science u n i v e r s i t y

B A Y L OR

route()

Happy mapping!!

david.kahle@gmail.com

http://sites.google.com/site/davidkahle/

Stamen Maps Watercolor