Question

I am visualizing a panel dataset with geom_point where y = var1, x = year, and color = var2. The problem is that there are many overlapping points, even with horizontal jitter.

Reducing the point size or setting a low alpha value is undesirable because both reduce the visual impact of the second variable, which has a very long right skew. I would like ggplot to place the points with the highest values of var2 on top of all other overlapping points.

Reproducible example:

df <- data.frame(diamonds)

ggplot(data = df,aes(x=factor(cut),y=carat,colour=price)) + 
  geom_point(position=position_jitter(width=.4))+
  scale_colour_gradientn(colours=c("grey20","orange","orange3"))

How does one place the points with highest values in df$price on top of an overlapping stack of points?

Answer 1

It looks as though grid plots in the order of the data,

library(grid)

d <- data.frame(x=c(0.5,0.52),y=c(0.6,0.6), fill=c("blue","red"),
                stringsAsFactors=FALSE)

grid.newpage()
with(d,grid.points(x,y,def='npc', pch=21,gp=gpar(cex=5, fill=fill)))
with(d[c(2,1),], grid.points(x,y-0.2,def='npc', pch=21,
                             gp=gpar(cex=5, fill=fill)))

so I would suggest you first reorder your data.frame, and pray that ggplot2 won’t mess with it 🙂

library(ggplot2)
library(plyr)
df <- diamonds[order(diamonds$price, decreasing=TRUE), ]
# alternative with plyr
df <- arrange(diamonds, desc(price))
last_plot() %+% df