R Graphics

Generating graphics for statistical analysis using R

Bo Werth
Statistician STI/EAS

Graphics in statistical software

programming

quality: reproducibility, calculations
efficiency gains: scaling
output flexibility: PDF, Word, HTML

interactive creation (e.g. Excel, Tableau)

development time
changing requirements
maintenance and transparency

`R` Graphics Systems

google trends: python, r, ggplot

traditional graphics

graphics facilities of the S language
fast, intuitive
create plot with high-level function
add elements with low-level functions

grid graphics system

produce complete plots
ideal proportions
facetting or multi-panel conditioning
trellis, lattice, ggplot

The organization of `R` graphics

R graphics system

Publication quality graphics with `ggplot`

implementation of Leland Wilkinson's "grammar of graphics" (2005)
independent components that can be composed in different ways
not limited to a set of pre-specified graphics
create new graphics that are precisely tailored

Interactive Graphics using JavaScript

ggvis: ggplot for dynamic charts based on vega.js
rCharts: interface to high-level js libraries building e.g. on d3.js
- highcharts, polyplot, nvd3, ricksaw for statistical charts
- crosslet, datamaps, leaflet etc. for map visualisations

traditional S-PLUS graphics

pen on paper model:
- can only draw on top of the plot
- cannot modify or delete existing content
no (user accessible) representation of the graphics
includes both tools for drawing primitives and entire plots
generally fast, but have limited scope

traditional S-PLUS graphics: `example(plot)`

plot of chunk plotexample

traditional S-PLUS graphics: plot regions

Single plot regions

Single Plot Regions

Multiple plot regions

Multi Plot Regions

traditional S-PLUS graphics: plot regions

op <- par(mfrow = c(2, 2),
          mar = c(3, 0, 0, 0))
plot(...); plot(...); plot(...); plot(...)
## At end of plotting, reset to previous settings:
par(op)

the documentation can be looked up with ?par()
margins are measured in multiples of lines of text
modifying traditional graphics state settings via par() has a persistent effect

traditional S-PLUS graphics: plot regions

op <- par(mfrow = c(2, 2),
          mar = c(3, 0, 0, 0))
plot(...); plot(...); plot(...); plot(...)
## At end of plotting, reset to previous settings:
par(op)

mfrow and mfcol control the number of figure regions on a page

par mfrow

par mfcol

traditional S-PLUS graphics: controlling plot regions

Graphics state settings controlling plot regions

diagram for controlling widths and horizontal locations
plot region = figure region - figure margins
plt: location of the plot region (l, r, b, t)
pin: size of the plot region, (width, height)
pty: m: use all available space, s: preserve square format

traditional S-PLUS graphics: colors and colours

colours()[1:4] # 657 color names

## [1] "white"         "aliceblue"     "antiquewhite"  "antiquewhite1"

col2rgb("transparent") # see the RGB values for a particular color name

##       [,1]
## red    255
## green  255
## blue   255

rgb(1, 0, 0) # Red-Green-Blue triplet of intensities, format #RRGGBB, FF = 255

## [1] "#FF0000"

traditional S-PLUS graphics: `pch` point symbols

pch symbols

A particular data symbol is selected by specifying an integer between 0 and 25 or a single character for the pch graphical setting. In the diagram, the relevant integer or character pch value is shown in grey to the left of the relevant symbol.

Source: http://vis.supstat.com/2013/04/plotting-symbols-and-color-palettes/

traditional S-PLUS graphics

example(barplot)

plot of chunk barplot

example(boxplot)

plot of chunk boxplotex

traditional S-PLUS graphics

example(pairs)

plot of chunk pairplot

example(persp)

plot of chunk perspplot

traditional S-PLUS graphics: `example(stars)`

plot of chunk starsegment

plot of chunk starradar

traditional S-PLUS graphics: `example(mosaicplot)`

plot of chunk mosaic1

plot of chunk mosaic2

traditional S-PLUS graphics: conditioning plot, `example(coplot)`

plot of chunk coplot1

plot of chunk coplot2

traditional S-PLUS graphics: lm example

plot of chunk lmplot

traditional S-PLUS graphics: Agglomerative Nesting (Hierarchical Clustering)

plot of chunk agnesplot

traditional S-PLUS graphics

plot(hclust(d = dist(USArrests), method = "average"), main=title)

plot of chunk unnamed-chunk-2

lattice / grid graphics: pre and post drawing

plot of chunk oztemp

draw map of Australia
draw average monthly temperatures for six cities

lattice / grid graphics: embedding plots in grid viewports

plot of chunk viewport

create dendrogram object and cut it into four subtrees
define lattice panel function to draw the dendrograms
make base plot region correspond to the created viewport
use traditional plot() function to draw the dendrogram

Layered Grammar of Graphics

A statistical graphic is a mapping from data to

geometric objects (points, lines, bars)
with aesthetic attributes (colour, shape, size)
in a coordinate system (cartesian, polar, map projection)

and optionally entails

statistical transformations of the data (binning, counting)
faceting to generate the same graphic for different subsets of the dataset

ggplot2 attempts to produce any kind of statistical graphic using

a compact syntax and independent components to facilitate extensions
the grid package to exercise low-level control over the appearance of the plot

Wickham, H. (2009). ggplot2. doi:10.1007/978-0-387-98141-3

Comparison `plot()` and `qplot()`

plot(x, y)

plot of chunk plot1

qplot(x, y)

plot of chunk qplot1

Comparison `plot()` and `qplot()`

plot(x, y, type = "l")

plot of chunk plot2

qplot(x, y, geom = "line")

plot of chunk qplot2

Comparison `plot()` and `qplot()`

plot(x, y, type = "s")

plot of chunk plot3

qplot(x, y, geom = "step")

plot of chunk qplot3

Comparison `plot()` and `qplot()`

plot(x, y, type = "b")

plot of chunk plot4

qplot(x, y, geom = c("point", "line"))

plot of chunk qplot4

`mtcars` dataset

Data from the 1974 Motor Trend US magazine for 32 automobiles (1973-74 models). The variables are the following:

mpg Miles/(US) gallon
cyl Number of cylinders
disp Displacement (cu.in.)
hp Gross horsepower
drat Rear axle ratio
wt Weight (lb/1000)
qsec 1/4 mile time
vs V/S
am Transmission (0 = automatic, 1 = manual)
gear Number of forward gears
carb Number of carburetors

Comparison `plot()` and `qplot()`

boxplot(wt~cyl,
        data=mtcars, col="lightgray")

plot of chunk boxplotcomp

qplot(factor(cyl), wt,
      data=mtcars, geom=c("boxplot", "jitter"))

plot of chunk boxqplotcomp

Comparison `plot()` and `qplot()`

hist(mtcars$wt)

plot of chunk histplotcomp

qplot(mtcars$wt, geom = "histogram",
      binwidth = 0.5, color = factor(0))

plot of chunk histqplotcomp

Comparison `plot()` and `qplot()`

cdplot(mtcars$wt, factor(mtcars$cyl))

plot of chunk cdplotcomp

qplot(mtcars$wt, fill=factor(mtcars$cyl),
      geom="density")

plot of chunk cdqplotcomp

`diamonds` dataset

A dataset containing the prices and other attributes of almost 54,000 diamonds. The variables are as follows:

price price in US dollars ($326-$18,823)
carat weight of the diamond (0.2-5.01)
cut quality of the cut (Fair, Good, Very Good, Premium, Ideal)
colour diamond colour, from J (worst) to D (best)
clarity a measurement of how clear the diamond is (I1 (worst), SI1, SI2, VS1, VS2, VVS1, VVS2, IF (best))
x length in mm (0-10.74)
y width in mm (0-58.9)
z depth in mm (0-31.8)
depth total depth percentage = z / mean(x, y) = 2 * z / (x + y) (43-79)
table width of top of diamond relative to widest point (43-95)

ggplot2 `qplot()`: single line of code

qplot(x=carat, y=price, colour=clarity,
  data=diamonds, geom=("point"))

plot of chunk qplotex1

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

plot of chunk qplotex2

ggplot2 `ggplot()`: add layers for more control using "`+`"

ggplot(data=diamonds) +
    geom_point(aes(x=price, y=carat, colour=color)) +
    facet_grid(. ~ clarity)

plot of chunk ggplotex

ggplot2 `ggplot()`: add layers for more control using "`+`"

ggplot(data = df,aes(x = x,y = y)) + # source: https://plot.ly/ggplot2/geom_errorbar/
    geom_errorbar(aes(ymin = ymin,ymax = ymax), colour = 'steelblue', width = 0.2) +
    geom_errorbarh(aes(xmin = xmin,xmax = xmax), colour = 'steelblue', height = 0.4) +
    geom_point(color = "black", size = 3)

plot of chunk ggplotlayer

ggplot Themes

theme_stata()

plot of chunk ggthemesstata

theme_economist()

plot of chunk ggthemeseconomist

ggplot Themes

theme_fivethirtyeight()

plot of chunk ggthemesfivethirtyeight

Tableau theme_igray()

plot of chunk ggthemesigray

ggplot Themes: `ggthemr`

pal.crN <- c('#95B3D7','#F79646','#8064A2','#4BACC6','#9BBB59','#C0504D')
ugly <- define_palette(
  swatch = pal.crN, gradient = c(lower = pal.crN[1L], upper = pal.crN[2L]))
ggthemr(ugly)
ggplot(dsamp, aes(x=price, fill=cut)) + geom_bar(binwidth = 500)

plot of chunk ggthemr

ggthemr_reset()

gridSVG: gapminder

gapminder animation

gridSVG animates a ggplot object before the output is flattened for export to a graphics device
the animation is obtained from mapping to the time dimension (annual since 1950)
size = population
color = continent
- blue: Europe
- red: Asia
- green: Africa
- yellow: America

gridSVG: R&D Expendidures

rnd animation

ggvis: `/demo/dynamic.r`

dynamic stacked bars

ggvis stacked bars

moving data points

ggvis moving points

rCharts: nvd3 Sparklines

p2 <- nPlot(uempmed ~ date, data = economics, type = 'sparklinePlus')
p2$chart(xTickFormat="#!function(d) {return d3.time.format('%b %Y')(new Date( d * 86400000 ));}!#")
p2$print('chart2')

rCharts: highcharts

p3 <- hPlot(Pulse ~ Height, data = MASS::survey, type = "bubble", title = "Zoom demo",
            subtitle = "bubble chart", size = "Age", group = "Exer")
p3$chart(zoomType = "xy"); p3$exporting(enabled = T); p3$print('chart3')

Reporting with `knitr`

use rmarkdown syntax
generate charts from data, e.g. using ggplot
include key figures in narrative, e.g. descriptive statistics
convert to various output formats (Word, PDF, HTML)

Reporting with `knitr`

knitr output

Reporting with `knitr`

# A Minimal Example for Markdown
This is a minimal example of using **knitr** to produce an _HTML_ page from _Markdown_.
## R code chunks
Now we write some code chunks in this markdown file:
'''{r computing}
x <- 1+1 # a simple calculator
set.seed(123)
rnorm(5)  # boring random numbers
'''
We can also produce plots:
'''{r graphics}
par(mar = c(4, 4, .1, .1))
with(mtcars, {
  plot(mpg~hp, pch=20, col='darkgray')
  lines(lowess(hp, mpg))
})
'''

Report Templates

brew: generate input files for knitr

apply function to country vector

create.report <- function(x, prepend = "report_icio_tiva_") {
  Rmd.file <- file.path(path.Rmd, paste0(prepend, x, ".Rmd"))
  rmd.file <- file.path(path.rmd, paste0(prepend, x, ".rmd")) # .md doesn't convert hash tags
  brew(file = file.path(path, "report_icio_tiva.brew"), output = Rmd.file)
  knit(input = Rmd.file, output = rmd.file)
  out.file <- paste0(prepend, x, ".rmd")
  return(out.file)
}
coulist <- c("AUT", "DEU", "ESP", "IRL", "USA")
results <- sapply(as.character(coulist), create.report)

`brew` template example

'''{r preamble, echo = FALSE}
cou <<- '<%= x %>'
country <- as.character(namereg$country[match(cou, namereg$cou)])
natnlty <- as.character(namereg$coupron[match(cou, namereg$cou)])
customtext <- cntext[,colnames(cntext)==cou]
'''
# Trade in Value-added: 'r country'
## EXGRDVA\_EX {#exgrdvaex}
### Domestic value added content of gross exports, 'r year', %
'''{r fig1, fig.path="figures/report_icio_tiva/<%= x %>/", fig.height=5, fig.width=10,
      echo=FALSE, message=FALSE}
    source(file.path(path, "code", "figure1.R"))
'''
'r country' domestic value-added content of its exports is, at 'r .perc1'%, 'r .rel1'
the OECD average in 'r year'.
'r if(!is.na(customtext[1])) customtext[1]'

Reporting Tools and Platforms

ReporteRs: create MS doc and ppt with editable graphics and formatted tables

Word logo

rapporter / pander: ruby on rails online platform for report generation

rapporter logo

jekyll: ruby on rails framework to generate static homepages from rmd files (i.e. evaluate R code with knitr, see http://10.101.26.220, http://r-pkgs.had.co.nz/ or http://adv-r.had.co.nz)

References

ggplot2 http://ggplot2.org/
R for Data Science http://r4ds.had.co.nz/
ggplot2 extensions https://www.ggplot2-exts.org/
static plotting showcase http://yutannihilation.github.io/allYourFigureAreBelongToUs/
interactive plotting showcase http://gallery.htmlwidgets.org/
CRAN Task View: Graphics http://cran.r-project.org/web/views/Graphics.html
Shiny R Graph Catalog http://shinyapps.stat.ubc.ca/r-graph-catalog/

R Graphics

Generating graphics for statistical analysis using R

Graphics in statistical software

R Graphics Systems

The organization of R graphics

Publication quality graphics with ggplot

Interactive Graphics using JavaScript

traditional S-PLUS graphics

traditional S-PLUS graphics: example(plot)

traditional S-PLUS graphics: plot regions

traditional S-PLUS graphics: plot regions

traditional S-PLUS graphics: plot regions

traditional S-PLUS graphics: controlling plot regions

traditional S-PLUS graphics: colors and colours

traditional S-PLUS graphics: pch point symbols

traditional S-PLUS graphics

traditional S-PLUS graphics

traditional S-PLUS graphics: example(stars)

traditional S-PLUS graphics: example(mosaicplot)

traditional S-PLUS graphics: conditioning plot, example(coplot)

traditional S-PLUS graphics: lm example

traditional S-PLUS graphics: Agglomerative Nesting (Hierarchical Clustering)

traditional S-PLUS graphics

lattice / grid graphics: pre and post drawing

lattice / grid graphics: embedding plots in grid viewports

Layered Grammar of Graphics

Comparison plot() and qplot()

Comparison plot() and qplot()

Comparison plot() and qplot()

Comparison plot() and qplot()

mtcars dataset

Comparison plot() and qplot()

Comparison plot() and qplot()

Comparison plot() and qplot()

diamonds dataset

ggplot2 qplot(): single line of code

ggplot2 ggplot(): add layers for more control using "+"

ggplot2 ggplot(): add layers for more control using "+"

ggplot Themes

ggplot Themes

ggplot Themes: ggthemr

gridSVG: gapminder

gridSVG: R&D Expendidures

ggvis: /demo/dynamic.r

dynamic stacked bars

moving data points

rCharts: nvd3 Sparklines

rCharts: highcharts

Reporting with knitr

Reporting with knitr

Reporting with knitr

Report Templates

brew template example

Reporting Tools and Platforms

References

`R` Graphics Systems

The organization of `R` graphics

Publication quality graphics with `ggplot`

traditional S-PLUS graphics: `example(plot)`

traditional S-PLUS graphics: `pch` point symbols

traditional S-PLUS graphics: `example(stars)`

traditional S-PLUS graphics: `example(mosaicplot)`

traditional S-PLUS graphics: conditioning plot, `example(coplot)`

Comparison `plot()` and `qplot()`

Comparison `plot()` and `qplot()`

Comparison `plot()` and `qplot()`

Comparison `plot()` and `qplot()`

`mtcars` dataset

Comparison `plot()` and `qplot()`

Comparison `plot()` and `qplot()`

Comparison `plot()` and `qplot()`

`diamonds` dataset

ggplot2 `qplot()`: single line of code

ggplot2 `ggplot()`: add layers for more control using "`+`"

ggplot2 `ggplot()`: add layers for more control using "`+`"

ggplot Themes: `ggthemr`

ggvis: `/demo/dynamic.r`

Reporting with `knitr`

Reporting with `knitr`

Reporting with `knitr`

`brew` template example