Introduction to the mountainplot package"


The mountainplot package provide an extension to the lattice package that allows for the consutruction of mountain plots, which are also known as folded empirical cumulative distribution plots.


opts_chunk$set(fig.align="center", fig.width=6, fig.height=6)

Load the package and use the singer data from the lattice package. Combine the first and second parts of each voice part into a new variable called section.

data(singer, package = "lattice")
parts <- within(singer, {
section <- voice.part
section <- gsub(" 1", "", section)
section <- gsub(" 2", "", section)
section <- factor(section)
# Change levels to logical ordering
levels(parts$section) <- c("Bass","Tenor","Alto","Soprano")

Mountain plot

A mountainplot, or folded empircal cumulative distribution function, is similar to an ordinary empirical CDF, but once the cumulative probability reaches 0.50, the CDF is inverted, decreasing back down instead of continuing upward.

Here is an example of the traditional empirical CDFs.

require(latticeExtra) # for ecdfplot
ecdfplot(~height|section, data = parts, groups=voice.part, type='l',
         main="Empirical CDF",
         auto.key=list(columns=4), as.table=TRUE)

Here is a view of the same data shown with a mountain plot.

mountainplot(~height|section, data = parts,
             groups=voice.part, type='l',
             main="Folded Empirical CDF",
             auto.key=list(columns=4), as.table=TRUE)

@monti1995folded suggests that a mountain plot is helpful with exploring data and makes it easier to:

  1. Determine the median.
  2. Determine the range.
  3. Determine central or tail percentiles of any specified value.
  4. Observe outliers.
  5. Observe unusual gaps in the data.
  6. Examine the data for symmetry.
  7. Compare multiple distributions.
  8. Visually examine the sample size.

Additionally, the area under the curve is equal to the mean absolute deviation (MAD) @xue2011pfolded.

Diabetic mice example

@huh1995exploring developed at the same time the concept of the flipped empirical distribution function. The following code creates a mountainplot of Hand's diabetic mice data, which can be compared to Huh's version.

dmice <- data.frame(
          82,100,98,150,243,68,228,131,73,18,20,100,72,133,465,40,46,34, 44),
  group=c(rep('normal',20), rep('alloxan', 18), rep('insulin', 19))
mountainplot(~albumen, data=dmice, group=group, auto.key=list(columns=3),
             main="Diabetic mice", xlab="Nitrogen-bound bovine serum albumen")

Session information



Try the mountainplot package in your browser

Any scripts or data that you put into this service are public.

mountainplot documentation built on July 13, 2017, 9:01 a.m.