I am finally learning
ggplot2 for elegant graphics. One of the first plots that I wanted to make was a length frequency histogram. As it turns out, there are a few “tricks” to make the histogram appear as I expect most fisheries folks would want it to appear – primarily, left-inclusive (i.e., 100 would be in the 100-110 bin and not the 90-100 bin). Below are length frequency histograms that I like.
The data I use are lengths of Lake Erie Walleye (Sander vitreus) captured during October-November, 2003-2014. These data are available in my
FSAdata package and formed ma of the examples in Chapter 12 of the Age and Growth of Fishes: Principles and Techniques book. My primary interest is in the
tl (total length in mm),
loc variables (see here for more details) and I will focus on 2010 (as an example).
Basic Length Frequency
Making the histogram begins by identifying the data.frame to use in
data= and the
tl variable to use for the
x-axis as an
ggplot(). The histogram is then constructed with
geom_hist(), which I customize as follows:
- Set the width of the length bins with
- By default the bins are centered on breaks created from
binwidth=. The bins can be changed to begin on these breaks by using
boundary=. The value that
boundary=, which is set to the beginning of a first break, regardless of whether that break is in the data or not. I use
boundary=0so that bins will start on breaks that make sense relative to
binwidth=(e.g., 0, 25, 50, 75, etc.).
- Bins are left-exclusive and right-inclusive by default, but including
closed="left"will make the bins the desired left-inclusive and right-exclusive.
- The fill color of the bins is set with
fill=(I prefer a slight gray).
- The outline color of the bins is set with
color=(defaults to the same as
fill=; I prefer a dark boundary to make the bins obvious).
scale_x_continuous() are primarily used to provide labels (i.e.,
names) for the y- and x-axes, respectively. By default, the bins of the histogram will “hover” slightly above the x-axis, which I find annoying. The
scale_y_continuous() is used to expand the lower limit of the y-axis by a
multiple of 0 (thus, not expand the lower-limit) and expand the upper limit of the y-axis by a
multiple of 0.05 (thus, the upper-limit will by 5% higher than the tallest bin so that the top frame of the plot will not touch the tallest bin). Finally,
theme_bw() gives a classic “black-and-white” feel to the plot (rather than the default plot with a gray background).
Note that the resultant plot was assigned to an object. Thus, the object name must be run to see the plot.
This base object/plot can also be modified by adding (using
+) to it as demonstrated later.
Bins Stacked by Another Variable
It may be useful to see the distribution of categories of fish (e.g., sex) within the length frequency bins. To do this, move the
geom_histogram() to an
geom_histogram() and set it equal to the variable that will identify the separation within each bin (e.g.,
sex). The bins will be stacked by this variable if
geom_histogram() (this is the default and would not need to be explicitly set below). The fill colors for each group can be set in a number of ways, but they are set manually below with
Stacked histograms are difficult to interpret in my opinion. In a future post, I will show how to use empirical density functions to examine distributions among categories. For the time being, see below.
Separated by Other Variable(s)
A strength of
ggplot2 is that it can easily make the same plot for several different levels of another variable; e.g., separate length frequency histograms by sex. The plot can be separated into different “facets” with
facet_wrap()m which takes the variable to separate by within
vars() as the first argument.
If the faceted groups have very different sample sizes then it may be useful to use a potentially different y-axis scale for each facet by including
facet_wrap(). Similarly, a potentially different scale can be used for each x-axis with
scales="free_x" or for both axes with
Plots may be faceted over multiple variables with
facet_grid(), where the variables that identify the rows and variables for a grid of facets are included (within
cols=, respectively. Both scales can not be “free” with
facet_grid() and the scale is only “free” within a row or column.
This post is likely not news to those of you that are familiar with
ggplot2. However, I am going to try to post some examples here as I learn
ggplot2 in hopes that hit will help others. This is the first of what I hope will be more frequent posts.
Other related (non-
ggplot2) posts are