{"id":886,"date":"2015-08-24T17:04:05","date_gmt":"2015-08-25T00:04:05","guid":{"rendered":"http:\/\/homepages.uc.edu\/~yaozo\/wordpress\/?p=886"},"modified":"2022-11-22T05:51:46","modified_gmt":"2022-11-22T05:51:46","slug":"taking-control-of-qualitative-colors-in-ggplot2","status":"publish","type":"post","link":"https:\/\/zhuoyao.net\/index.php\/2015\/08\/24\/taking-control-of-qualitative-colors-in-ggplot2\/","title":{"rendered":"Taking control of qualitative colors in ggplot2"},"content":{"rendered":"

Optional getting started advice<\/a><\/h3>\n

Ignore if you don’t need this bit of support.<\/em><\/p>\n

This is one in a series of tutorials in which we explore basic data import, exploration and much more using data from the Gapminder project<\/a>. Now is the time to make sure you are working in the appropriate directory on your computer, perhaps through the use of anRStudio project<\/a>. To ensure a clean slate, you may wish to clean out your workspace and restart R (both available from the RStudio Session menu, among other methods). Confirm that the new R process has the desired working directory, for example, with thegetwd()<\/code> command or by glancing at the top of RStudio’s Console pane.<\/p>\n

Open a new R script (in RStudio, File > New > R Script). Develop and run your code from there (recommended) or periodicially copy “good” commands from the history. In due course, save this script with a name ending in .r or .R, containing no spaces or other funny stuff, and evoking “ggplot2” and “colors”.<\/p>\n

Load the Gapminder data and ggplot2<\/code><\/a><\/h3>\n

Assuming the data can be found in the current working directory, this works:<\/p>\n

gDat <- read.delim<\/span>(\"gapminderDataFiveYear.txt\"<\/span>)<\/code><\/pre>\n
Plan B (I use here, because of where the source of this tutorial lives):<\/p>\n
## data import from URL\ngdURL <- \"http:\/\/www.stat.ubc.ca\/~jenny\/notOcto\/STAT545A\/examples\/gapminder\/data\/gapminderDataFiveYear.txt\"<\/span>\ngDat <- read.delim<\/span>(file =<\/span> gdURL)<\/code><\/pre>\n
Basic sanity check that the import has gone well:<\/p>\n
str<\/span>(gDat)<\/code><\/pre>\n
## 'data.frame':    1704 obs. of  6 variables:\n##  $ country  : Factor w\/ 142 levels \"Afghanistan\",..: 1 1 1 1 1 1 1 1 1 1 ..\n##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...\n##  $ pop      : num  8425333 9240934 10267083 11537966 13079460 ...\n##  $ continent: Factor w\/ 5 levels \"Africa\",\"Americas\",..: 3 3 3 3 3 3 3 3 ..\n##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...\n##  $ gdpPercap: num  779 821 853 836 740 ...<\/code><\/pre>\n
Drop Oceania, which only has two continents<\/p>\n
## drop Oceania\njDat <- droplevels<\/span>(subset<\/span>(gDat, continent != \"Oceania\"<\/span>))\nstr<\/span>(jDat)<\/code><\/pre>\n
## 'data.frame':    1680 obs. of  6 variables:\n##  $ country  : Factor w\/ 140 levels \"Afghanistan\",..: 1 1 1 1 1 1 1 1 1 1 ..\n##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...\n##  $ pop      : num  8425333 9240934 10267083 11537966 13079460 ...\n##  $ continent: Factor w\/ 4 levels \"Africa\",\"Americas\",..: 3 3 3 3 3 3 3 3 ..\n##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...\n##  $ gdpPercap: num  779 821 853 836 740 ...<\/code><\/pre>\n
Load ggplot2<\/code>:<\/p>\n
library<\/span>(ggplot2)<\/code><\/pre>\n
Take control of the size and color of points<\/a><\/h3>\n
Let’s use ggplot2<\/code> to move towards the classic Gapminder bubble chart. Crawl then walk then run.<\/p>\n
First, make a simple scatterplot for a single year.<\/p>\n
jYear <- 2007<\/span>\nq <- ggplot<\/span>(subset<\/span>(jDat, year == jYear),\n            aes<\/span>(x =<\/span> gdpPercap, y =<\/span> lifeExp)) + scale_x_log10<\/span>()\nq + geom_point<\/span>()<\/code><\/pre>\n
\"\"<\/div>\n
Take control of the plotting symbol, its size, and its color. Use obnoxious settings so that success versus failure is completely obvious. Now is not the time for the delicate operation of inserting your fancy color scheme. Be bold!<\/p>\n
## do I have control of size and fill color? YES!\nq + geom_point<\/span>(pch =<\/span> 21<\/span>, size =<\/span> 8<\/span>, fill =<\/span> I<\/span>(\"darkorchid1\"<\/span>))<\/code><\/pre>\n
\"\"<\/div>\n
Circle area = population<\/a><\/h3>\n
We want the size of the circle to reflect population. Since we have direct control of the radius, we invert the relation a<\/span>r<\/span>e<\/span>a<\/span>=<\/span>\u03c0<\/span>r<\/span>2<\/span><\/span><\/span><\/span><\/span><\/span> to determine the point size from the country’s population. I have two complaints with my first attempt: the circles are still too small for my taste and I don’t want the size legend. So in my second attempt, I suppress the legend with show_guide = FALSE<\/code> and I increase the range of sizes by explicitly setting the range for the scale that maps (<\/span><\/span>\u221a<\/span>p<\/span>o<\/span>p<\/span>\/<\/span><\/span><\/span>\u03c0<\/span>)<\/span><\/span><\/span><\/span><\/span> into circle size.<\/p>\n
q + geom_point<\/span>(aes<\/span>(size =<\/span> sqrt<\/span>(pop\/pi)), pch =<\/span> 21<\/span>)\n(r <- q +\n   geom_point<\/span>(aes<\/span>(size =<\/span> sqrt<\/span>(pop\/pi)), pch =<\/span> 21<\/span>, show_guide =<\/span> FALSE<\/span>) +\n   scale_size_continuous<\/span>(range=<\/span>c<\/span>(1<\/span>,40<\/span>)))<\/code><\/pre>\n
\"\"\"\"<\/p>\n
Circle fill color determined by a factor<\/a><\/h3>\n
Now I use aes()<\/code> to map a factor to color. For the moment, I settle for the continent<\/code> factor and for the automatic color scheme. I also facet by continent. Why? Because it will be helpful below for checking my progress on using my custom color scheme. Since all the countries, say, in Europe, are some shade of green, if the continent facets have circles of many colors, I’ll know something’s wrong.<\/p>\n
(r <- r + facet_wrap<\/span>(~ continent))\nr + aes<\/span>(fill =<\/span> continent)<\/code><\/pre>\n
\"\"\"\"<\/p>\n
Get the color scheme for the countries<\/a><\/h3>\n
Elsewhere, I devised a color scheme for the Gapminder countries. We will not discuss it’s construction here, but will merely pull it off the web. You can view it in PDF form here<\/a>.<\/p>\n
## get the country color scheme\ngdURL <- \"http:\/\/www.stat.ubc.ca\/~jenny\/notOcto\/STAT545A\/examples\/gapminder\/data\/gapminderCountryColors.txt\"<\/span>\ncountryColors <- read.delim<\/span>(file =<\/span> gdURL, as.is =<\/span> 3<\/span>) # protect color<\/span>\nstr<\/span>(countryColors)<\/code><\/pre>\n
## 'data.frame':    142 obs. of  3 variables:\n##  $ continent: Factor w\/ 5 levels \"Africa\",\"Americas\",..: 1 1 1 1 1 1 1 1 ..\n##  $ country  : Factor w\/ 142 levels \"Afghanistan\",..: 95 39 43 28 118 121 ..\n##  $ color    : chr  \"#7F3B08\" \"#833D07\" \"#873F07\" \"#8B4107\" ...<\/code><\/pre>\n
head<\/span>(countryColors)<\/code><\/pre>\n
##   continent          country   color\n## 1    Africa          Nigeria #7F3B08\n## 2    Africa            Egypt #833D07\n## 3    Africa         Ethiopia #873F07\n## 4    Africa Congo, Dem. Rep. #8B4107\n## 5    Africa     South Africa #8F4407\n## 6    Africa            Sudan #934607<\/code><\/pre>\n
The data.frame countryColors<\/code> has one row per country and three variables: country, continent, and color. The color variable holds the RGB hex strings encoding the color scheme.<\/p>\n
Note:<\/strong> The row order of countryColors<\/code> is not alphabetical. The countries are actually sorted by size (in which particular year, I don’t recall) within continent, reflecting the logic by which the scheme was created. No problem. Ideally, nothing in your analysis should depend on row order, although that’s not always possible in reality.<\/p>\n
Prepare the color scheme for use with ggplot2<\/code><\/a><\/h3>\n
In the Grammar of Graphics, a scale<\/strong> controls the mapping from a variable in the data to an aesthetic. So far we’ve let the coloring \/ filling scale be determined automatically by ggplot2<\/code>. But to use our custom color scheme, we need to take control of the mapping of the country<\/code> factor into fill color in geom_point()<\/code>.<\/p>\n
We will use scale_fill_manual<\/code>, a member of a family of functions for customization of the discrete scales. The main argument isvalues =<\/code>, which is a vector of aesthetic values — fill colors, in our case. If this vector has names, they will be consulted during the mapping. This is incredibly useful! Below, we isolate the vector of hex strings providing the country colors and give this vector the country names as names. This saves us from any worry about the order of levels of the country<\/code> factor, the row order of the data, or exactly which countries are being plotted.<\/p>\n
jColors <- countryColors$color\nnames<\/span>(jColors) <- countryColors$country\nhead<\/span>(jColors)<\/code><\/pre>\n
##          Nigeria            Egypt         Ethiopia Congo, Dem. Rep. \n##        \"#7F3B08\"        \"#833D07\"        \"#873F07\"        \"#8B4107\" \n##     South Africa            Sudan \n##        \"#8F4407\"        \"#934607\"<\/code><\/pre>\n
Make the ggplot2<\/code> bubble chart<\/a><\/h3>\n
This is deceptively simple at this point. Like many things, it looks really easy, once we figure everything out! The last two bits we add are to use aes()<\/code> to specify that the country should be mapped to color and to use scale_fill_manual()<\/code> to specify our custom color scheme.<\/p>\n
r + aes<\/span>(fill =<\/span> country) + scale_fill_manual<\/span>(values =<\/span> jColors)<\/code><\/pre>\n
\"\"<\/div>\n
Epilogue: re-make the plot to reveal small countries<\/a><\/h3>\n
We know from earlier work with lattice<\/code> that large countries can effectively hide the data from small countries, by covering them up. This is a case where, sadly, the row order of the data truly affects the visual output. ggplot2<\/code> is no less vulnerable to this thanlattice<\/code> or base graphics here. So, to get closure, we sort the data on year and then on population and remake the plot, revealing all the code.<\/p>\n
jDat <- jDat[with<\/span>(jDat, order<\/span>(year, -1<\/span> * pop)), ]\nggplot<\/span>(subset<\/span>(jDat, year == jYear),\n       aes<\/span>(x =<\/span> gdpPercap, y =<\/span> lifeExp)) + scale_x_log10<\/span>() +\n  geom_point<\/span>(aes<\/span>(size =<\/span> sqrt<\/span>(pop\/pi)), pch =<\/span> 21<\/span>, show_guide =<\/span> FALSE<\/span>) +\n  scale_size_continuous<\/span>(range=<\/span>c<\/span>(1<\/span>,40<\/span>)) +\n  facet_wrap<\/span>(~ continent) +\n  aes<\/span>(fill =<\/span> country) + scale_fill_manual<\/span>(values =<\/span> jColors)<\/code><\/pre>\n
\"\"<\/div>\n
References<\/a><\/h3>\n