Improve poor automatic tick position choices without explicitly specifying breaks

Question

I find that ggplot2 sometimes produces too few tick marks when using scale_y_log10. I am trying to produce plots automatically from arbitrary data, and I'm looking for a way to increase the number of tick marks without explicitly specifying them (since I don't know ahead of time what the data will be). For instance, here's a function to create a simple scatterplot with a log-y-scale:

example_plot <- function(x) {
  p <- ggplot(d, aes(x=MW, y=rel.Ki)) + 
    geom_point() +
    scale_y_log10()
  p
}

This will often work well, but with the following data

d <- structure(list(MW = c(89.09, 174.2, 147.13, 75.07, 131.17, 131.17, 146.19, 149.21, 165.19, 115.13, 181.19, 117.15), rel.Ki = c(2.91438577473767, 1, 1.07761254731238, 1.0475715900998, 0.960123906592881, 0.480428471483881,  1.50210548081627, 0.318457530434953, 0.458477212731015, 1.92246139937586,  0.604121577795352, 2.4111345825694)), .Names = c("MW", "rel.Ki"), class = "data.frame", row.names = c(1L, 6L, 11L, 16L, 21L, 26L, 31L, 36L, 41L, 47L, 54L, 59L))

it produces

print(example_plot(d))

The single tick mark on the y axis is not very helpful. Is there any way I can prevent this situation, short of rewriting the automatic tick-position-picking function?

Answer 1

You can set the limits programmatically. For example, using the data you provide, we can define the limits in the function like this:

example_plot <- function(x){
  # identify the range of data
  lims <- c(10^floor(log10(min(x$rel.Ki, na.rm=TRUE))), 
    10^ceiling(log10(max(x$rel.Ki, na.rm=TRUE))))
  # require ggplot2
  require('ggplot2')
  # create the plot
  p <- ggplot(data = x, aes(x = MW, y = rel.Ki)) + 
    geom_point() +
    scale_y_log10(limits = lims)
  p
}

print(example_plot(d))

Then you get a plot with ticks at the nearest decade:

Then, if you want to add a logarithmic grid, use the breaks option to scale_y_log10() as Marius et al. suggest:

 example_plot <- function(x){
  # identify the range of data      
  lims <- c(10^floor(log10(min(x$rel.Ki, na.rm=TRUE))), 
            10^ceiling(log10(max(x$rel.Ki, na.rm=TRUE))))  
   # require ggplot2
  require('ggplot2')
  # create the plot
  p <- ggplot(data = x, aes(x = MW, y = rel.Ki)) + 
    geom_point() +
    scale_y_log10(breaks = pretty(x = lims, n = 5),
                  limits = lims) 
  p 
}

print(example_plot(d))

Personally I prefer logarithmic plots to show at least an order of magnitude variation, so this approach helps ensure that happens.

Answer 2

An interesting discovery I just made by reading ?continuous_scale is that the breaks argument can be:

a function, that when called with a single argument, a character vector giving the limits of the scale, returns a character vector specifying which breaks to display.

So to guarantee a certain number of breaks, you could do something like:

break_setter = function(lims) {
  return(seq(from=as.numeric(lims[1]), to=as.numeric(lims[2]), length.out=5))
}

ggplot(d, aes(x=MW, y=rel.Ki)) + 
    geom_point() +
    scale_y_log10(breaks=break_setter)

Obviously the very simple example function is not very well adapted to the logarithmic nature of the data, but it does show how you could approach this a bit more programmatically.

---

You can also use pretty, which takes a suggestion for a number of breaks and returns nice round numbers. Using

break_setter = function(lims) {
    return(pretty(x = as.numeric(lims), n = 5))
}

yields the following:

Even better, we can make break_setter() return an appropriate function with whatever n you want and a default of, say, 5.

break_setter = function(n = 5) {
   function(lims) {pretty(x = as.numeric(lims), n = n)}
}

ggplot(d, aes(x=MW, y=rel.Ki)) + 
    geom_point() +
    scale_y_log10(breaks=break_setter())  ## 5 breaks as above

ggplot(d, aes(x=MW, y=rel.Ki)) + 
    geom_point() +
    scale_y_log10(breaks=break_setter(20))