ggplot2: easy graphics with R

Transcription

1 ggplot2: easy graphics with R Ilmari Ahonen Department of Mathematics and Statistics University of Turku ilmari.ahonen@utu.fi ggplot2 1 / 26

2 Versatile and rich graphics are widely concidered a major strength of the R language. However, achieving pleasing results with the base installation of R often still requires: a lot of time extensive knowledge of the various plotting functions hacking ggplot2 package by prof. Hadley Wickham aims to simplify the process and bring everything under one common language ilmari.ahonen@utu.fi ggplot2 2 / 26

3 The philosophy of ggplot2, based on the Grammar of Graphics 1 by Leland Wilkinsons, separetes plot layers into distinct categories: data and aesthetic mapping (x, y, color, size,...) statistical transformation (log, mean, histogram, smoother,...) a geometric object (dot, bar, line, box,...) position adjustment (jitter, stack,...) 1 Wilkinson L, 5, The Grammar of Graphics (Statistics and Computing), Springer, ISBN: ilmari.ahonen@utu.fi ggplot2 3 / 26

4 ggplot2 installs and loads like any other R package: # to install: install.packages("ggplot2") library(ggplot2) ilmari.ahonen@utu.fi ggplot2 4 / 26

5 We ll use the mtcars data as an example: cyl disp drat wt qsec vs am gear carb Mazda RX Mazda RX4 Wag Datsun Hornet 4 Drive Hornet Sportabout Valiant Duster Merc 240D Merc Merc Merc 280C Merc 450SE Merc 450SL Merc 450SLC Cadillac Fleetwood Lincoln Continental Chrysler Imperial Fiat Honda Civic Toyota Corolla ilmari.ahonen@utu.fi ggplot2 5 / 26

6 Let s investigate the relationship between the number horse powers and fuel consumption. We ll start by setting these variables as aesthetics: my.plot <- ggplot(data=mtcars, aes(x=, y=)) Now we have just an empty plot frame where we still need to add the geometric attributes. ilmari.ahonen@utu.fi ggplot2 6 / 26

7 my.plot + geom_point() ilmari.ahonen@utu.fi ggplot2 7 / 26

8 Many other geometrics are also grammatically correct my.plot + geom_point() + geom_line() ilmari.ahonen@utu.fi ggplot2 8 / 26

9 my.plot + geom_bar(stat="identity") ilmari.ahonen@utu.fi ggplot2 9 / 26

10 my.plot + geom_text(label=rownames(mtcars), size=4) Maserati Bora Ford Pantera L Camaro Duster Z28360 Chrysler Imperial ln Continental lac Fleetwood Merc Merc 450SLC Merc Hornet 450SE Pontiac 450SL Ferrari Sportabout Firebird Dino Dodge AMC Challenger Javelin Merc 280C Merc 280 Valiant Mazda Hornet Volvo RX4 RX4 4142E Drive Wag Lotus Europa Toyota Datsun Merc Corona Porsche Merc 240DFiat X1 9 Fiat Toyota 128 Coro Honda Civic ilmari.ahonen@utu.fi ggplot2 10 / 26

11 Let s go with the dot plot for the next examples p <- my.plot + geom_point() ilmari.ahonen@utu.fi ggplot2 11 / 26

12 It is easy to add additional aesthetichs to existing plots p + aes(size=cyl) cyl ilmari.ahonen@utu.fi ggplot2 12 / 26

13 p + aes(size=cyl, col=factor(gear)) cyl factor(gear) ilmari.ahonen@utu.fi ggplot2 13 / 26

14 Every type of plot can be faceted p + aes(size=cyl, col=factor(gear)) + facet_grid(. ~ am) 0 1 cyl factor(gear) ilmari.ahonen@utu.fi ggplot2 14 / 26

15 Adding some typical analysis results, such as errorbars or smoothers, is straightforward p + geom_smooth(method=lm) 0 ilmari.ahonen@utu.fi ggplot2 15 / 26

16 p + geom_smooth(method=loess) 0 ilmari.ahonen@utu.fi ggplot2 16 / 26

17 Again, additional aesthetics can be added p + geom_smooth(method=lm) + aes(col=factor(gear)) 400 factor(gear) ilmari.ahonen@utu.fi ggplot2 17 / 26

18 The appearance of the plots can be modified using themes. Let s continue with the previous plot p2 <- p + geom_smooth(method=lm) + aes(col=factor(gear)) p2 400 factor(gear) ilmari.ahonen@utu.fi ggplot2 18 / 26

19 p2 + theme_classic() 400 factor(gear) ilmari.ahonen@utu.fi ggplot2 19 / 26

20 p2 + theme_minimal() 400 factor(gear) ilmari.ahonen@utu.fi ggplot2 20 / 26

21 Now let s create our own theme our.awesome.theme <- theme( axis.text = element_text(size = 14), legend.key = element_rect(fill = "navy"), legend.background = element_rect(fill = "lightgreen"), legend.position = c(0.1, 0.9), panel.grid.major = element_line(colour = "lightgray"), panel.background = element_rect(fill = "beige") ) ilmari.ahonen@utu.fi ggplot2 21 / 26

22 p2 + our.awesome.theme 400 factor(gear) ilmari.ahonen@utu.fi ggplot2 22 / 26

23 p2 + theme_minimal() + scale_colour_grey() + aes(linetype=factor(gear), shape=factor(gear)) 400 factor(gear) ilmari.ahonen@utu.fi ggplot2 23 / 26

24 The package documentation and additional resources can be found on the project web page: ggplot2 24 / 26

25 Some additional points: qplot() behaves similarly to the base R s plot() and can be used for creating simple graphics quickly without using the full potential of the grammar ggplot2 is not suited for interactive graphics and is often too slow for smooth animations Some commonly used plots, such as scatter plot matrices are difficult to express using the grammar and are thus not straightforwad to produce with ggplot2. See GGally package for help. ilmari.ahonen@utu.fi ggplot2 25 / 26

26 Some additional resources that were used for this presentation: Rstudio: IDE for R and more knitr: embed R analyses and results in tex documents ilmari.ahonen@utu.fi ggplot2 26 / 26