ggplot2: stat_summary()

## knitr configuration: http://yihui.name/knitr/options#chunk_options
opts_chunk$set(comment = "", error= TRUE, warning = FALSE, message = FALSE,
               tidy = FALSE, cache = F, echo = T,
               fig.width = 6, fig.height = 6)

## R configuration
options(width = 116, scipen = 5)

References

• ggplot2 book: http://ggplot2.org/book/
• Help topics: http://docs.ggplot2.org/current/

Load ggplot2

library(ggplot2)

Summary

Summarizing the y values for each unique x value can be useful.

Currently, there is a bug (version 0.9.3) that prevents summary stat from working with line geom.

http://stackoverflow.com/questions/13775150/unusual-behavior-of-ggplot2-0-9-3

Even the ggplot2 web site has gone wrong.

http://docs.ggplot2.org/current/stat_summary.html

Summary function(s) should be given via these options

fun.data: Complete summary function. Should take data frame as input
          and return data frame as output

fun.ymin: ymin summary function (should take numeric vector and return
          single number)

   fun.y: y summary function (should take numeric vector and return
          single number)

fun.ymax: ymax summary function (should take numeric vector and return
          single number)

Functions to get multiple summary values

Wrap up a selection of summary functions from Hmisc to make it easy to use with 'stat_summary'.

Description:
     See the Hmisc documentation for details of their options.

Usage:
       mean_cl_boot(x, ...)
       mean_cl_normal(x, ...)
       mean_sdl(x, ...)
       median_hilow(x, ...)

Continuous X values

p.movies2 <- ggplot(data = movies,
                    mapping = aes(x = year, y = rating))

## median and line geom
p.movies2 + layer(geom = "line",
                  stat = "summary",
                  fun.y = median)

## median_hilow and smooth geom
p.movies2 + layer(geom = "smooth",
                  stat = "summary",
                  fun.data = median_hilow)

## mean and line geom
p.movies2 + layer(geom = "line",
                  stat = "summary",
                  fun.y = mean)

## mean_cl_boot and smooth geom
p.movies2 + layer(geom = "smooth",
                  stat = "summary",
                  fun.data = mean_cl_boot)

Discrete X values

Now the crossbar geom is having a problem.

p.movies3 <- ggplot(data = movies,
                    mapping = aes(x = round(rating), y = log10(votes)))

## mean and point geom
p.movies3 + layer(geom = "point",
                  stat = "summary",
                  fun.y = mean)

## mean and line geom
p.movies3 + layer(geom = "line",
                  stat = "summary",
                  fun.y = mean)

## mean_cl_normal and errorbar geom
p.movies3 + layer(geom = "errorbar",
                  stat = "summary",
                  fun.data = mean_cl_normal)

## median_hilow and pointrange geom
p.movies3 + layer(geom = "pointrange",
                  stat = "summary",
                  fun.data = median_hilow)

## median_hilow and crossbar geom
p.movies3 + layer(geom = "crossbar",
                  stat = "summary",
                  fun.data = median_hilow)

## min/mean/max and crossbar geom
p.movies3 + layer(geom = "crossbar",
                  stat = "summary",
                  fun.y = mean, fun.ymax = max, fun.ymin = min)

Individual summary functions

midm <- function(x) mean(x, trim = 0.5)

p.movies3 +
    layer(stat = "summary",
          geom = "point",
          mapping = aes(color = "trimmed"),
          fun.y = midm
          ) +
    layer(stat = "summary",
          geom = "point",
          mapping = aes(color = "raw"),
          fun.y = mean
          ) +
    scale_color_discrete(name = "Mean")

Single summary function with multiple return values

iqr <- function(x, ...) {
    qs <- quantile(as.numeric(x), probs = c(0.25, 0.75), na.rm = TRUE)
    names(qs) <- c("ymin","ymax")
    qs
}

p.movies2 + layer(stat = "summary",
                  geom = "ribbon",
                  fun.data = iqr)