ggplot2: Adding a geom without affecting limits

Question

I would like to add additional geoms to a ggplot density plot, but without changing the displayed limits of the data and without having to compute the desired limits by custom code. To give an example:

set.seed(12345)
N = 1000
d = data.frame(measured = ifelse(rbernoulli(N, 0.5), rpois(N, 100), rpois(N,1)))
d$fit = dgeom(d$measured, 0.6)
ggplot(d, aes(x = measured)) + geom_density() + geom_line(aes(y = fit), color = "blue")

ggplot(d, aes(x = measured)) + geom_density() + geom_line(aes(y = fit), color = "blue") + coord_cartesian(ylim = c(0,0.025))

In the first plot, the fit curve (which fits the "measured" data quite badly) obscures the shape of the measured data: I would like to crop the plot to include all data from the first geom, but crop the fit curve, as in the second plot:

While I can produce the second plot with coord_cartesian, this has two disadvantages:

1. I have to compute the limits by my own code (which is cumbersome and error-prone)
2. Computing the limits by my own code is not compatible with faceting. It is not possible (AFAIK) to provide per-facet axis limits with coord_cartesian. I however need to combine the plot with facet_wrap(scales = "free")

The desired output would be achieved, if the second geom was not considered when computing coordinate limits - is that possible without computing the limits in custom R code?

The question R: How do I use coord_cartesian on facet_grid with free-ranging axis is related, but does not have a satisfactory answer.

Answer 1

One thing you could try is to scale fit and use geom_density(aes(y = ..scaled..)

Scaling fit between 0 and 1:

d$fit_scaled <- (d$fit  - min(d$fit)) / (max(d$fit) - min(d$fit))

Use fit_scaled and ..scaled..:

ggplot(d, aes(x = measured)) + 
  geom_density(aes(y = ..scaled..)) + 
  geom_line(aes(y = fit_scaled), color = "blue")

This can be combined with facet_wrap():

d$group <- rep(letters[1:2], 500) #fake group

ggplot(d, aes(x = measured)) + 
  geom_density(aes(y = ..scaled..)) + 
  geom_line(aes(y = fit_scaled), color = "blue") + 
  facet_wrap(~ group, scales = "free")

An option that does not scale the data:

You can use the function multiplot() from http://www.cookbook-r.com/Graphs/Multiple_graphs_on_one_page_(ggplot2)/

multiplot <- function(..., plotlist=NULL, file, cols=1, layout=NULL) {
  library(grid)
  plots <- c(list(...), plotlist)

  numPlots = length(plots)

  if (is.null(layout)) {

    layout <- matrix(seq(1, cols * ceiling(numPlots/cols)),
                    ncol = cols, nrow = ceiling(numPlots/cols))
  }

 if (numPlots==1) {
    print(plots[[1]])

  } else {

    grid.newpage()
    pushViewport(viewport(layout = grid.layout(nrow(layout), ncol(layout))))

    for (i in 1:numPlots) {

      matchidx <- as.data.frame(which(layout == i, arr.ind = TRUE))

      print(plots[[i]], vp = viewport(layout.pos.row = matchidx$row,
                                      layout.pos.col = matchidx$col))
    }
  }
}

With this function you can combine the two plots, which makes it easier to read them:

multiplot(
  ggplot(d, aes(x = measured)) + 
    geom_density() +
    facet_wrap(~ group, scales = "free"),
  ggplot(d, aes(x = measured)) +  
    geom_line(aes(y = fit), color = "blue") + 
    facet_wrap(~ group, scales = "free")
)

This will give you:

And if you want to compare groups next to each other, you can use facet_grid() instead of facet_wrap() with cols = 2 in multiplot():

multiplot(
  ggplot(d, aes(x = measured)) + 
    geom_density() +
    facet_grid(group ~ ., scales = "free"),
  ggplot(d, aes(x = measured)) +  
    geom_line(aes(y = fit), color = "blue") + 
    facet_grid(group ~ ., scales = "free"),
  cols = 2
)

And it looks like this:

Answer 2

You can try to calculate the max y-limit first. Then plot.

d1 <- d %>% 
  mutate(max_dens=round(max(density(measured)$y), 2))

ggplot(d1, aes(x=measured)) + 
   geom_line(aes(y=fit), color = "blue") +
   geom_density() + 
   coord_cartesian(ylim = c(0, unique(d1$max_dens)))