Graphics I Or Making things pretty in R.

Transcription

1 Graphics I Or Making things pretty in R rebecca.smith@kcl.ac.uk

2 In this session See the range of options for graphics in R Be able to use basic graphics Make clear, attractive graphs Highlight some useful graphics packages

3 Weight Better than Excel? Chick Weights horsebean linseed soybean sunflower meatmeal casein 50 0 horsebean linseed soybean sunflower meatmeal casein Feed

4 Benefits of R?

5 Benefits of R? R is great when you want to produce one type of graph very frequently in a consistent format (e.g. common bar graphs, scatter plots) Producing high resolution plots for publication and presentations Making composite plots Having control over small details (making it perfect) Producing clear, beautiful plots! The best plots are those which viewers immediately understand! Limit clutter, colour, and extraneous objects

6 Examples demo(graphics) demo(persp) library(lattice) demo(lattice) Data to play with data(chickwts) data(trees) data(sunspot.year) data(occupationalstatus) data(cars)

7 Visualising Data data(chickweight) weight Time Chick Diet Matrix of plots of columns of a data frame (everything by everything) plot(chickweight)

8 Plot Types Scatter plot: plot() Bar graph: barplot() Strip charts: stripchart() Box plot: boxplot() Pie chart: pie() Histogram: hist() Volcano plot: drawvolcanoplot [requires madb library] Survival curve: survfit followed by plot [requires survival library] You can easily make Volcano plots, QQ plots, or ROC curves with the plot function; however, using specific functions can save time

9 Plot Types hist(chickweight$weight) boxplot(weight~time) stripchart(weight~time, vertical=t)

10 Useful additions to scripts Add a title main= title X and Y label ylab= weight, xlab= group Change the point style and size pch=19, cex=1.5 Line type and width lty=2, lwd=2 Change the colour of plotted parts col=c( red, white, blue ) More later and at

11 pch and lty

12 Adding lines abline() Adds one or more straight lines through the current plot Needs Intercept and slope Uses all par() features like lty, lwd and col Can use results directly from a linear model plot(weight~time, data=chickweight) abline(coef(lm(chickweight$ weight~chickweight$time)) [1], coef(lm(chickweight$weigh t~chickweight$time))[2], lwd=2, lty=2, col="red")

13 Axes boxplot(weight~time), xaxt="n") axis(1, 1:12, letters[0:12]) boxplot(weight~time), xaxt="n") axis(1, c(2,4,6,8,10,12), c("b", "d", "f", "g", "i","k"))

14 Legend Adding a legend to a plot outside of the script Can tell position, group names, etc. fill is for boxes with colours in col is for line and point colour Need to specify lty or pch legend(4, 40, c( Group1", Group2"), fill=c("red", "blue")) legend(4, 40, c( Group1", Group2"), col=c("red", "blue"), pch=1)

15 Exporting your creations Image type PDF pdf( test.pdf ) TIFF tiff( test.tiff ) PNG png( test.png ) Bitmap bmp( test.bmp ) JPEG jpeg( test.jpeg ) Window size and units (px, mm, in or cm) width = 480, height = 480, units = "px" Resolution res = NA Point size pointsize = 12 tiff( image.tiff, width = 25, height = 25, units = cm, res = 100, pointsize = 10) plot(x~y) dev.off()

16 Be aware... tiff("cwres.tiff") plot(weight~time, data=chickweight) dev.off() tiff("cwres1.tiff", width = 25, height = 25, units = "cm", res = 100, pointsize = 10) plot(weight~time, data=chickweight) dev.off()

17 Margins You can adjust how much of a margin around the plot you give Can adjust all margins individually left right v v par(mar=c(2,2,1,1)) ^ ^ bottom top Also allow plotting in the margins mtext(text, side = 3) <- Text is written in one of the four margins par(xpd=na)<- If FALSE, all plotting is clipped to the plot region, if TRUE, all plotting is clipped to the figure region, and if NA, all plotting is clipped to the device region

18 Matrix Make a matrix of multiple plots par(mfrow=c(2,2)) <- 2 by 2 matrix par(mfrow=c(3,1)) <- 1 column, 3 rows One figure in row 1 and two figures in row 2 layout(matrix(c(1,1,2,3), 2, 2, byrow = TRUE))

19 Don t reinvent the wheel... Many graphics packages available in R If there is something complicated, chances someone have made a package or written a script you can adapt ggplot (Jack) sciplot Circos WGCNA

20 sciplot Useful package for one-way or higher experimental designs Produces line plots and bargraphs library(sciplot) bargraph.ci lineplot.ci

21 In Action bargraph.ci(time, weight, group=diet, data=chickweight, legend=true, x.leg=0, ylim=c(0,300), col=rainbow(4, s = 1, v = 1))

22 More complex Can easily use R to create plots for data visualisation and perform QC of data Useful for Showing principal components QQ-plots Sex typing Manhattan plots No matter what you use to plot Keep it simple and understandable Eliminate distractions Chose best fit graph