wwl - 1 year ago 115
R Question

# floating.pie in plotrix in R

I'm using R. My ultimate objective is to have a GIS map of Minnesota at the county level, and in each county there'll be a pie chart, representing the proportion of primary, secondary, and tertiary people employed.

I have a dataframe containing employment statistics. Each row is a county, and the columns are employment in the primary, secondary, and tertiary sector. The first ten rows look like this.

``````      [,1]   [,2]   [,3]
[1,]    0     94   2208
[2,]    0  45568  65678
[3,]    0   3124   6262
[4,]   48   3908  11278
[5,]    0   6949   9779
[6,]    0    283    992
[7,]    0   7797  27130
[8,]   15   4471   6710
[9,]   51   1973   5768
[10,]   44  14188  23040
``````

To put a pie chart, I use
`floating.pie`
function of
`plotrix`
package and iterate over all counties. (
`mn_county`
is my GIS shapefile, but you don't have to bother about that)

``````floating.pie(coordinates(mn_county)[s, ][1], coordinates(mn_county)[s, ][2], c(pri, sec,tert), radius = rad
, col = c(rgb(255, 0, 0, max = 255, alpha = 125, names = "red50"), rgb(0, 255, 0, max = 255, alpha = 125, names = "green50"), rgb(0, 0, 255, max = 255, alpha = 125, names = "blue50")))
``````

For my pie chart, I want primary employment to be red, secondary employment to be green, and tertiary employment to be blue. The problem is, when primary employment = 0, R shades secondary employment as red and tertiary employment as green. How can I prevent this from happening?

In principle you could hack the `floating.pie` function yourself (zeros are excluded because they break some of the internal computations; it should be possible to work around this) or send a request to the package maintainer ...

An easy hack is to add a small value to all of your categories (small enough that you'll never notice the difference visually); alternatively you could just use `min(pri,eps)` ...

``````dat <- data.frame(pri=c(0,10,20),
sec=c(10,20,30),
tert=c(10,10,10))
n <- 3
library(plotrix)
plot(c(0,n+1),c(0,n+1),axes=FALSE,ann=FALSE,type="n")
eps <- 0.001
for (i in 1:n) {
with(dat[i,],floating.pie(i,i,c(pri+eps,sec+eps,tert+eps),
• red/green/blue aren't the best possible colour representations (primary red/green are indistinguishable to dichromats), you might look at the `RColorBrewer` package ...