Intro to ggplot2 - Sheffield R Users Group, Feb 2015
2 likes1,865 views
The document provides an introduction to ggplot2, a powerful R package for creating 2D graphics based on the grammar of graphics theory. It covers essential concepts such as layers, aesthetics, and statistical transformations, demonstrating how to produce various visualizations like scatterplots, boxplots, and histograms using the ggplot2 functions. Additionally, it emphasizes the importance of using long format data and provides resources for further learning.
Intro to ggplot2 - Sheffield R Users Group, Feb 2015
- 1. Introduction to ggplot22 Paul Richards, ScHARR, The University of Sheffield Thursday, February 26, 2015
- 2. Introduction ggplot2 is a package written by Hadley Wickham Powerful but easy to use functions for 2D graphics Based on the “Grammar of Graphics” theory by Leland Wilkinson use install.packages() to install latest version
- 3. Concepts ggplot2 works with data.frames aesthetic = a feature we can see on the graphic (shape, size, colour etc) map from data to aestheics (i.e. different colour per group) layer = geometric object + data/aesthetics + statistical transformation a graphic will also have scale(s) and a co-ordinate system may also have facets - subsetting plot by some characteristic(s)
- 4. Quick note on “plot” The “plot” function in base R is just a wrapper for lots of methods Behaviour depends on what object is supplied Can require some manual tinkering to get it to work as required Example, using the “iris” data, plot petal length vs sepal length Different colour for each species Point size varies by sepal width
- 5. Scatterplot with “plot” with(iris, plot(Sepal.Length, Petal.Length, col = Species, cex = Sepal.Width)) 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 1234567 Sepal.Length Petal.Length
- 6. Problems Cannot specify dataset within function No legend, have to add manually via legend() which is fiddly to use Arguments are a bit “dumb” - we need to rescale Sepal.Width to get better point sizes Need to use different functions to add new geometric objects, e.g. regression lines
- 7. Qplot qplot(Sepal.Length, Petal.Length, data = iris, color = Species, size = Sepal.Width) 2 4 6 5 6 7 8 Sepal.Length Petal.Length Species setosa versicolor virginica Sepal.Width 2.0 2.5 3.0 3.5 4.0
- 8. Qplot() Qplot is the nearest equivalent to plot() in ggplot2 For single layer plots this is easy enough to use Use geom argument to change plot type
- 9. Boxplot example qplot(Species, Sepal.Length, data = iris, geom="boxplot") 5 6 7 8 setosa versicolor virginica Species Sepal.Length
- 10. Violin example qplot(Species, Sepal.Length, data = iris, geom="violin") 5 6 7 8 setosa versicolor virginica Species Sepal.Length
- 11. Histogram example qplot(Sepal.Length, data = iris, fill = Species) 0 5 10 4 5 6 7 8 Sepal.Length count Species setosa versicolor virginica
- 12. ggplot() For multilayer plots or where more flexibility is required ggplot() sets up the default data and aesthetic mappings add layers using the “+” operator and appropriate functions all aesthetic mappings are wrapped in aes() function global changes (e.g. set all points to “red”) go outside
- 13. Iris scatterplot again ggplot(data = iris, aes(x = Sepal.Length, y = Petal.Length, color = Species)) + geom_point(aes(size = Sepal.Width)) 2 4 6 5 6 7 8 Sepal.Length Petal.Length Species setosa versicolor virginica Sepal.Width 2.0 2.5 3.0 3.5 4.0
- 14. Alternative ggplot(data = iris, aes(x = Sepal.Length, y = Petal.Length, color = Species)) + geom_point(size = 5) 2 4 6 5 6 7 8 Sepal.Length Petal.Length Species setosa versicolor virginica
- 15. Adding to plots ggplots are objects so you can save them as you go You can then add new layers etc to the saved object gg1 <- ggplot(data = iris, aes(x = Sepal.Length, y = Petal.Length, color = Species)) + geom_point(aes(size = Sepal.Width)) gg2 <- gg1 + geom_smooth() gg2
- 16. Iris plot with loess smoother 2 4 6 5 6 7 8 Sepal.Length Petal.Length Species setosa versicolor virginica Sepal.Width 2.0 2.5 3.0 3.5 4.0
- 17. Add contours gg3 <- gg2 + geom_density2d() gg3
- 18. Add contours 2 4 6 5 6 7 8 Sepal.Length Petal.Length Species setosa versicolor virginica Sepal.Width 2.0 2.5 3.0 3.5 4.0
- 19. Note on statistical transformations The loess smoother and contours in the previous plot are not part of data In base plot we would have to calculate them first In ggplot2 the stat_*() functions do such tranformations for us Often the corresponding geom_*() function does this automatically For more flexibility, use the stat_*() function and specify the geom you need
- 20. Histogram example using stat_bin stat_bin performs a 1d “binning” transformation i.e. a histogram transformation ggplot(data = iris, aes(x = Sepal.Length, fill = Species)) + stat_bin(binwidth=1,position="dodge")
- 21. Histogram example using stat_bin 0 10 20 30 4 6 8 Sepal.Length count Species setosa versicolor virginica
- 22. Density plot with facet Use facet_grid() to subset plots by up to 2 variables Function takes a formula as its main argument Row variable on left, column variable on right Use . if no variable needed ggplot(data = iris, aes(x = Sepal.Length)) + geom_density(fill = Species) + facet_grid(Species ~ .)
- 23. Density plot with facet 0.0 0.4 0.8 1.2 0.0 0.4 0.8 1.2 0.0 0.4 0.8 1.2 setosaversicolorvirginica 5 6 7 8 Sepal.Length density Species setosa versicolor virginica
- 24. Scale + axis control Adding title labels, axes etc is similar to adding layers Use the “+” operator with the appropriate functions ggplot(USArrests, aes(x=Assault,y=Murder)) + geom_point(aes(size=UrbanPop)) + labs(title = "Violent Crime Rates by US State, 1973", x = "Arrests for Assault (per 100 000)", y = "Arrests for Murder (per 100 000)")
- 25. Scale + axis control 0 5 10 15 100 200 300 Arrests for Assault (per 100 000) ArrestsforMurder(per100000) UrbanPop 40 50 60 70 80 90 Violent Crime Rates by US State, 1973
- 26. Logged x axis ggplot(USArrests, aes(x=Assault,y=Murder)) + geom_point(aes(size=UrbanPop)) + labs(title = "Violent Crime Rates by US State, 1973", x = "Arrests for Assault (per 100 000)", y = "Arrests for Murder (per 100 000)") + scale_x_log10()
- 27. Logged x axis 0 5 10 15 100 Arrests for Assault (per 100 000) ArrestsforMurder(per100000) UrbanPop 40 50 60 70 80 90 Violent Crime Rates by US State, 1973
- 28. Conclusion ggplot2 works best with “long” format data One row per observation, rather than different obs in different columns See “reshape2” package for easy conversion between “wide” and “long” data formats
- 29. Where to learn more Web documentation is a good place to start http://docs.ggplot2.org Lots of examples on blogs, stackoverflow etc. We have only scratched the surface here! Why not bring some example data visualisations to the next meeting? Tweet your plots @Sheffield_R_