Question

I create a simple scatter plot with ggplot2 and visualise the z-variable with a color:

 require(ggplot2)

 data = data.frame(x=runif(1000), y=runif(1000), vis=rf(1000, df1=1, df2=3))
 qplot(x=x, y=y, data=data, color=vis)

however, this is of course not very informative since the distribution is heavily skewed:

 hist(data$vis)

the problem – in my opinion – is the equidistant breaks that creates bins for data that is simply not in the sample.

so here is my question: is there an efficient way of overcoming this problem and creating more breaks where more data is available. or in other words i’m looking for non-linear breaks or non-equidistant braks.

Answer 1

Edit: probably something more similar to what you want:

breaks <- quantile(data$vis)
qplot(x=x, y=y, data = data, color = vis) + 
     scale_colour_gradientn(breaks = as.vector(breaks), colours = 
     c("grey", "blue", "red"), values = as.vector(breaks), 
     oob = identity, rescaler = function(x,...) x, labels = names(breaks))

Old answer: In this case breaks are not what you really want

qplot(x=x, y=y, data=data, color=vis) + scale_colour_gradient(breaks = 1:10 * 10)

Considering amount of data we have

quantile(data$vis, seq(0, 1, 0.1))
          0%          10%          20%          30%          40% 
9.294095e-07 1.883887e-02 8.059213e-02 1.646752e-01 3.580304e-01 
         50%          60%          70%          80%          90% 
6.055612e-01 9.463869e-01 1.638687e+00 2.686160e+00 5.308239e+00 
        100% 
1.693077e+02 

so possibly something like

qplot(x=x, y=y, data=data, color=vis) + scale_colour_gradient(limits = c(0,5))

would be good, here points > 5 are grey. A more complicated solution, which you maybe wanted in the first place would be this.