How can I fill the background with a color scale in function of a lineal model?

问题

I'm analysing an example of datavis in "The Functional Art " of Alberto Cairo (I recomend to you)

In that book, there are this example

And I try in R. In the bottom left graph (scatter plot)

I use the data from the book, and i calculate the efectives militaries in function of population with lm(efect ~ pop) and the budget in function of population + efect

And here is the question: How can I paint the background of the scatterplot with colorscale (gradient) where the color represent the value of BUDGET?

After that I want to put the points (population, efects) with the color of the budget. In this way, I can compare the budget of the country with the stimated budget in function of population + efect

I know basic R, I have installed ggplot2 and scales packages

I want a graph similar to

but with the correct colors.

Thank you.

回答1:

To get the background you can use geom_raster to get the effect you want. I'll use the same data set as suggested by @GeekOnAcid:

First get the data and fit a regression model:

crime = read.csv("http://datasets.flowingdata.com/crimeRatesByState2005.tsv", 
              header=TRUE, sep="\t")
 ##Fit the regression model
 m = lm(crime$burglary ~ crime$murder)

Next we create a grid for the background colour:

##Create a grid for the background colour
x = seq(1, 10, length.out=100)
y = seq(400, 1200, length.out=100)
z = expand.grid(x,y)

Now we need a distance measure for the gradient colour. I just used squared distance from the regression line:

z$grad = (z[,2] - (398.3 +  62.2*z[,1]))^2

Then plot:

require(ggplot2)    
ggplot(z) + geom_raster(aes(Var1, Var2, fill=grad)) + 
    geom_point(data=crime[1:15,], aes(murder, burglary, size=population),pch=1 ) + 
    geom_text(data=crime[1:15,], 
              aes(murder, burglary, label=state), 
              hjust=-0.2, size=4) + 
    scale_size_continuous(range=c(1,10)) + 
    scale_fill_continuous(high="red", low="white", trans="sqrt") +
    xlab("Murder") + ylab("Burglary")  + 
    guides(size=FALSE, fill=FALSE) +
    scale_y_continuous(expand=c(0, 0)) + 
    scale_x_continuous(expand=c(0, 0))

To get:

回答2:

To get you started, take a look at this tutorial by Nathan Yau, who shows how to create bubble chart with basic graphics in R. I only retrieved his solution to make it more relevant for your example.

#get some example data
crime <- read.csv("http://datasets.flowingdata.com/crimeRatesByState2005.tsv", 
                  header=TRUE, sep="\t")
#define the radius of circles
radius <- sqrt( crime$population/ pi )
#makes your plot, bg defines colour, inches scales circles
symbols(crime$murder[1:15], crime$burglary[1:15], circles=radius[1:15], 
        inches=0.5, bg=gray(0.9), xlab="Murder Rate", ylab="Burglary Rate")
#makes your labels, you can offset them by adding values to x and y arguments
text(crime$murder[1:15], crime$burglary[1:15], crime$state[1:15], cex=0.6)

The background thing will be tricky, but are you sure you really need it? Even with different colours it will look ugly...

来源：https://stackoverflow.com/questions/13221830/how-can-i-fill-the-background-with-a-color-scale-in-function-of-a-lineal-model

标签

colors

ggplot2

scale

data-visualization