¡¡¡¡¡¡ Freedom !!!!!!

SPSS

ArcGIS

Notepad

GWR4.0

Notepad

ArcGIS

Word

# Load data
data(LondonHP)
# Create distance matrix
DM <- gw.dist(dp.locat=coordinates(londonhp))
# Create optimized bandwidth
bw1 <- bw.gwr(PURCHASE~FLOORSZ, data=londonhp, kernel = "gaussian",dMat=DM)

## Fixed bandwidth: 28008.52 CV score: 901202470969 
## Fixed bandwidth: 17313.68 CV score: 842907727581 
## Fixed bandwidth: 10703.9 CV score: 736181883398 
## Fixed bandwidth: 6618.837 CV score: 607130814353 
## Fixed bandwidth: 4094.128 CV score: 529769270141 
## Fixed bandwidth: 2533.772 CV score: 496244493691 
## Fixed bandwidth: 1569.419 CV score: 558268461315 
## Fixed bandwidth: 3129.775 CV score: 504912379213 
## Fixed bandwidth: 2165.422 CV score: 500148436808 
## Fixed bandwidth: 2761.425 CV score: 498237171785 
## Fixed bandwidth: 2393.075 CV score: 496399101924 
## Fixed bandwidth: 2620.728 CV score: 496732595196 
## Fixed bandwidth: 2480.03 CV score: 496150911653 
## Fixed bandwidth: 2446.816 CV score: 496183570335 
## Fixed bandwidth: 2500.558 CV score: 496166149711 
## Fixed bandwidth: 2467.344 CV score: 496154814854 
## Fixed bandwidth: 2487.871 CV score: 496153635785 
## Fixed bandwidth: 2475.184 CV score: 496151178429 
## Fixed bandwidth: 2483.025 CV score: 496151494463 
## Fixed bandwidth: 2478.179 CV score: 496150836405 
## Fixed bandwidth: 2477.035 CV score: 496150899230 
## Fixed bandwidth: 2478.886 CV score: 496150839373 
## Fixed bandwidth: 2477.742 CV score: 496150850527 
## Fixed bandwidth: 2478.449 CV score: 496150833774 
## Fixed bandwidth: 2478.616 CV score: 496150834475 
## Fixed bandwidth: 2478.346 CV score: 496150834229 
## Fixed bandwidth: 2478.513 CV score: 496150833832 
## Fixed bandwidth: 2478.41 CV score: 496150833868 
## Fixed bandwidth: 2478.474 CV score: 496150833765 
## Fixed bandwidth: 2478.489 CV score: 496150833779 
## Fixed bandwidth: 2478.465 CV score: 496150833764 
## Fixed bandwidth: 2478.459 CV score: 496150833766 
## Fixed bandwidth: 2478.468 CV score: 496150833764 
## Fixed bandwidth: 2478.47 CV score: 496150833764 
## Fixed bandwidth: 2478.467 CV score: 496150833764 
## Fixed bandwidth: 2478.466 CV score: 496150833764 
## Fixed bandwidth: 2478.467 CV score: 496150833764 
## Fixed bandwidth: 2478.468 CV score: 496150833764 
## Fixed bandwidth: 2478.467 CV score: 496150833764 
## Fixed bandwidth: 2478.467 CV score: 496150833764 
## Fixed bandwidth: 2478.467 CV score: 496150833764 
## Fixed bandwidth: 2478.467 CV score: 496150833764 
## Fixed bandwidth: 2478.467 CV score: 496150833764

# Run GWR
gwr.res1 <- gwr.basic(PURCHASE~FLOORSZ, data=londonhp, 
                      bw=bw1,kernel = "gaussian", dMat=DM)

# Make map interactive
tmap_mode("view")
# Take spatial data, change format, make sure projection is correct, create map
gwr.res1$SDF %>% st_as_sf() %>% st_set_crs(27700) %>% 
  tm_shape()+tm_dots(col="FLOORSZ", palette="RdYlGn") -> plot1

• R is brilliant
• It made my life a lot easier
• Coming up next:
  • A very very short introduction, covering some of the basices
  • Too much material for the time we have
  • Goal is to (hopefully) enthuse and show what’s possible
  • Links to (free!) textbooks and resources at the end

By the end of this session you will be able to:

• Install packages, open packages
• Open data
• Create simple tables
• Create simple models
• Create maps with data and models

• Stop doing repetitive tasks

Wiersma et al. 2022

• Do reproducible science
  • Quantitative
  • And very much qualitative as well
• Create presentations

• Write entire papers

Rijnks et al. 2022

• R does everything better than any other software package

1. Overview of RStudio

2. Basic R commands

   • Installing and loading packages
   • Loading data
   • Viewing data
   • Brief overview of data types
   • Simple models
   • Find help
   • Learn to love errors

1. Links to books / examples

   • Book on geographic data science with R
   • Book on text mining with R
   • Some fun and useful links

• Stand-alone
• In IDE (e.g. RStudio)

• REPL: Instantaneous, but fleeting
• Quarto: bit more faff to write, but saves as document
• REPL: Could do comments using #
• Quarto: Text / presentation / notes / descriptions separate
• REPL: Model output, but barely legible
• Quarto: Lovely tables and figures

1. File –> New Project

2. New Directory –> New Project

3. Give it a name (e.g. Winterschool) and location (e.g. Desktop)

4. Start a new file in that project: Quarto Document

5. Sort of optional: Change the “editor” from “visual” to “source”

• install.packages in REPL
  • once installed you have them accessible on your pc
• load packages in Quarto
  • that way you can load only project-specific packages
• How to make R chunk in Quarto?

# Install the packages
install.packages("tidyverse")
install.packages("sf")
install.packages("tmap")
install.packages("GWmodel")
install.packages("RColorBrewer")
install.packages("kableExtra")
install.packages("gridExtra")
# Load the packages (note the lack of quotation marks)

library(tidyverse)
library(sf)
library(tmap)
library(GWmodel)
library(RColorBrewer)
library(kableExtra)

# Data as part of packages
data(LondonHP)
# Change from SpatialPointsDataFrame (old) to SF (modern)
london_sf <- londonhp %>% st_as_sf() %>% st_set_crs(27700)

# CSV files
#read.csv("path_to_data")

# Dutch CSV files (if your data comes in as: however many observations of 1 variable...)
#read.csv2("path_to_data")

# R proprietary files
#load("name_of_data.Rda")

# SPSS files? See package: "foreign" or "haven". Stata13 files? see package "readstata13"

# Shapefiles? st_read("path_to_shape.shp")

• Typing “<-” means object on left “gets” whatever is on the right
• Typing “->” vice versa
• Convention: “<-”
• Try it in REPL:
  • test <- 5
  • 2 * test
  • test_squared <- test * test
  • sqrt(test_squared)

1. “head” displays the first few rows of the data file

head(london_sf)

## Simple feature collection with 6 features and 20 fields
## Geometry type: POINT
## Dimension:     XY
## Bounding box:  xmin: 531900 ymin: 159400 xmax: 535700 ymax: 161700
## Projected CRS: OSGB36 / British National Grid
##   PURCHASE FLOORSZ TYPEDETCH TPSEMIDTCH TYPETRRD TYPEBNGLW TYPEFLAT BLDPWW1
## 0   157000      77         1          0        0         1        0       0
## 1   113500      75         0          0        1         0        0       1
## 2    81750      64         0          0        0         0        1       0
## 3   150000      95         0          1        0         0        0       0
## 4   190000     107         1          0        0         0        0       0
## 5   159950     100         0          1        0         0        0       0
##   BLDPOSTW BLD60S BLD70S BLD80S BLD90S BATH2 BEDS2 GARAGE1 CENTHEAT   UNEMPLOY
## 0        0      0      0      0      0     0     1       0        1 0.03566768
## 1        0      0      0      0      0     0     1       0        1 0.03566768
## 2        0      0      0      1      0     0     0       1        0 0.03566768
## 3        0      0      0      0      0     0     1       0        1 0.03566768
## 4        0      0      0      0      0     0     1       1        1 0.03566768
## 5        0      0      1      0      0     0     1       1        1 0.02408854
##        PROF BLDINTW              geometry
## 0 0.4786992       1 POINT (533200 159400)
## 1 0.4786992       0 POINT (533300 159700)
## 2 0.4786992       0 POINT (532000 159800)
## 3 0.4786992       1 POINT (531900 160100)
## 4 0.4786992       1 POINT (532800 160300)
## 5 0.4773715       0 POINT (535700 161700)

1. “View” (mind the capitalization) shows file

#View(london_sf)

• Bring up data documentation

?LondonHP

1. Load data
2. How many:
   • Garages
   • Post-war
   • From the 90s
3. Correlate
   • Price x Professional / managerial
   • Floorsize x unemployed
4. Plot map of purchase price

- use the dollar sign to access variables:
london_sf$BEDS2
- table(x) creates simple tables
- cor(x,y) correlates variables
- cor.test(x,y) runs correlation test
- draw map: 
qtm(shapefilename, dots.col="VARIABLE_YOU_WANT")
- note the quotation marks

london_sf %>% mutate(log_purch = log(PURCHASE)) %>%
tm_shape() +
  tm_dots(col="log_purch", palette="-RdYlGn")

# to see more options go to:
# vignette("tmap-getstarted")

• Model is an abstraction / simplification of reality
• All models are wrong \[therefore\] the scientist must be alert to what is importantly wrong
• The standard intro into modelling:
  • Correlation
  • ANOVA
  • Linear regression
  • And their expansions / non-parametric equivalents
• Simple, usually wrong, superseded by more complicated models

• Boston Marathon: very difficult to get into
• Qualifying time: age + gender stratified
  • Example: 18-34 years, men: 3.00, women: 3.30
• Bib numbers are sequentially with qualifying time
• You’d expect finish time to be linked to qual time
• Model: \[Time_{finish} = \alpha + \beta * Time_{quali} + \epsilon\]

marathoninvestigation.com, 17-2-2017

london_sf %>% st_drop_geometry() %>%
  ggplot(aes(x=FLOORSZ, y=PURCHASE))+
           geom_point() + theme_classic()

• Regression equation \[ y = \alpha + \beta_1 * x_1 + \beta_2 * x_2 + \epsilon\]
• In R: lm(y ~ x1 + x2, data=name_of_data)
• Be sure to save the output as an object!
  • lm1 <- lm(y ~ x1 + x2, data=name_of_data)
• Now you can have fun with it:
  • summary(lm1)
  • plot(lm1)
  • or in Quarto: xtable()

london_sf %>% st_drop_geometry() %>% 
  mutate(PURCHASE = log(PURCHASE), FLOORSZ = log(FLOORSZ)) %>%
  lm(PURCHASE ~ FLOORSZ, data=.) -> lm1
library(xtable)
# make sure to mark the chunk with  results='asis'
print(xtable(summary(lm1)),type='html')

1. Starting from the log-log model
2. Build a parsimonious model for house-prices
   • Be aware of dummy variables: what is your reference category
   • Make sure to save your models
   • For model selection (which one is best)
     • AIC()
3. Interpret the model

• Getting help in R is easy:
  • Package or function documentation: ?name_of_function
  • Online: cran is the official R website
  • Online: most packages will have a github repo
• Help outside R:
  • Stackexchange
  • r-bloggers
  • bookdowns

• Denoted in red in the console
• Try and read them fully and understand them
• They are meant to help
• You’ll develop a love-hate relationship with them

• Key book for learning geographic data-science

https://geocompr.robinlovelace.net

• Other great books:

  • https://uk.sagepub.com/en-gb/eur/an-introduction-to-r-for-spatial-analysis-and-mapping/book258267

  • https://r4ds.had.co.nz

• Text mining tweets
• Twitter still has an open API for data science
• Can be used for downloading selections of tweets
• Can give really interesting (or really wrong) insights

All of the following is from:

• http://varianceexplained.org/r/trump-tweets/
• This website gives code and interpretation
• It’s fascinating work, and typical of R:
  • Open, shared, reproducible

varianceexplained.org/r/trumptweets

varianceexplained.org/r/trumptweets

varianceexplained.org/r/trumptweets

varianceexplained.org/r/trumptweets

• Authoring books with R-Markdown (Yihui Xie)

https://bookdown.org/yihui/bookdown/

• Shiny apps (Hadley Wickham)

https://mastering-shiny.org

• GGPlot graphics (Hadley Wickham)

https://ggplot2-book.org

• General fun and interesting

https://r-bloggers.com

https://www.bigbookofr.com/index.html

“Reproducible science helps you do science, better science”

• Reproducing science is a valid way of “doing” science

• Reproducible science reduces errors in new studies

• Reproducible science reduces publication bias, effect size bias, etc.

• Reproducible science helps you do science

• Better science

• Reproducing science: Kuhn: The structure of Scientific Revolutions

• Open
  • Access
  • Source
• Reproducible

It’s also a great way of getting cited…

https://ejd.econ.mathematik.uni-ulm.de

• If you try R and want to discuss
• Get in touch
  • with me (r.h.rijnks@rug.nl)
  • or with whoever published the package you are working on. They are usually happy to help