Visualizing the spread of a numeric variable(s)
Histograms
Density plots
Ridgeline plots
Box plots
Violin plots
Examples show the distribution of penguin body masses (g) for Adelie, Chinstrap & Gentoo penguins.
The Santa Barbara Coastal Long Term Ecolgical Research (SBC LTER) site was established in 2000 to understand the ecology of coastal kelp forest ecosystems. A number of coastal rocky reef sites are outfitted with instrumentation that collect long-term monitoring data.
We’ll be exploring bottom temperatures recorded at Mohawk Reef, a near-shore rocky reef and one of the Santa Barbara Coastal (SBC) LTER research sites.
What are they?
- Histograms are used to represent the distribution of a numeric variable(s), which is cut into several bins. The number of observations per bin is represented by the height of the bar.
Need:
- a numeric variable with lots of values
- meaningful differences between values
Important considerations:
- bin width (30 bins by default)
- too few / too many bins
Twelve groups (month_name) is too many groups – especially when the range of temperature values for each of our groups largely overlap:
If you want to plot all groups, consider splitting them into small multiples. If so, does color add any valuable information? Remove if not:
Let’s instead compare just three months: April (generally the coldest month), October (generally a hot month), June (somewhere in between):
Use fill to fill bars with a specified color(s) and color to outline bars with a specified color(s):
mko_clean |>
mutate(month_name = factor(month_name, levels = month.name)) |>
filter(month_name %in% c("April", "June", "October")) |>
ggplot(aes(x = Temp_bot, fill = month_name)) +
geom_histogram(position = "identity", alpha = 0.5, color = "black") +
scale_fill_manual(values = c("#2C5374", "#ADD8E6", "#8B3A3A"))
Modify binwidth (30 bins by default) – does a bin width of 1 (degree Celsius) actually make sense? Consider scale of interest. Also be mindful when using bins – too few bins will result in loss of distribution shape.
mko_clean |>
filter(month_name %in% c("April", "June", "October")) |>
mutate(month_name = factor(month_name, levels = month.name)) |>
ggplot(aes(x = Temp_bot, fill = month_name)) +
geom_histogram(position = "identity", alpha = 0.5, binwidth = 1) +
scale_fill_manual(values = c("#2C5374", "#ADD8E6", "#8B3A3A"))
What are they?
- A smoothed version of a histogram. Density plots are representations of the distribution of a numeric variable(s), which uses a kernel density estimate (KDE) to show the probability density function of the variable. The area under each curve is equal to 1. Use a density plot when you are most concerned with the shape of the distribution.
Need:
- a numeric variable with lots of values
Important considerations:
- useful when you want to visualize the shape of your data (not affected by bin number)
- does not indicate sample size
- can be misleading with small data sets
- band width, which affects level of smoothing
Similar to the histogram, twelve groups (month_name) is too many groups! Consider small multiples (using facet_wrap()) if you want to keep all groups.
If you want to plot all groups, consider splitting them into small multiples. If so, does color add any valuable information? Remove if not:
Let’s instead compare three months: April (generally the coldest month), October (generally a hot month), June (somewhere in between):
Modify bandwidth by declaring a multiplier of the default bandwidth adjustment. Reducing the adjust argument reduces the amount of smoothing (default adjust = 1):
It’s easy to see that group B has a larger sample size than group A when looking at our histogram. Additionally, we can get a good sense of our data distribution. But what happens when you reduce the number of bins (e.g. set bins = 4)?
We lose information about sample size in our density plot (note that both curves are ~the same height, despite group B having 2x as many observations). However, they’re great for visualizing the shape of our distributions since they are unaffected by the number of bins.
Overlaying a histogram and density plot requires scaling down the histogram to match the density curve scale. Adding y = after_stat(density) within the aes() function rescales the histogram counts so that bar areas integrate to 1:
In a normal density plot, the area under the curve(s) is equal to 1. In a scaled density plot, the area under the curve reflects the number of observations for each group.
We can use scaled density plots to compare individual group distributions to the total distribution. We’ll do so using the palmerpenguins::penguins data set.
# use `after_stat(count)` to plot density of observations ----
ggplot(penguins, aes(x = body_mass_g, y = after_stat(count))) +
# plot full distribution curve with label "all penguins"; remove 'species' col so that this doesn't get faceted later on ----
geom_density(data = select(penguins, -species),
aes(fill = "all penguins"), color = "transparent") +
# plot second curve with label "species" ----
geom_density(aes(fill = "species"), color = "transparent") +
# facet wrap by species ----
facet_wrap(~species, nrow = 1) +
# update colors, x-axis label, legend position ----
scale_fill_manual(values = c("grey","#0C8346"), name = NULL) +
labs(x = "Body Mass (g)") +
theme(legend.position = "top")
What are they?
- Ridgeline plots show the distribution of a numeric variable for multiple groups.
Need:
- a numeric variable with lots of values
Important considerations:
- work best when you have > 6 groups
- works well when there is a clear pattern in the result (e.g. if there is an obvious ranking in groups) and / or when visualizing changes in distributions over time or space
The {ggridges} package has a number of different geoms for creating ridgeline plots that work well for data sets with larger group numbers (e.g. months). Two great geoms to explore (to start):
geom_density_ridges() to create a basic ridgeline plot:
Order by month (ideal, since months have an inherent order):
Order by mean or median (makes more sense when you have unordered groups):
mko_clean |>
mutate(month_name = fct_reorder(month_name, Temp_bot, .fun = mean)) |>
ggplot(mko_clean, mapping = aes(x = Temp_bot, y = month_name, fill = after_stat(x))) +
ggridges::geom_density_ridges_gradient() +
scale_fill_gradientn(colors = c("#2C5374","#849BB4", "#D9E7EC", "#EF8080", "#8B3A3A"))
rel_min_height adjusts trailing tails and scale controls the extent to which the different densities overlap)
Include a median line by using the stat_density_ridges() geom and setting the number of quantiles to 2:
Visualize the raw data underlying the density ridges (since our temperature data is too large (>473,000 rows), so we’ll use the palmerpenguins::penguins data set to demo):
Jittered points
What are they?
- Box plots summarize the distribution of a numeric variable for one or several groups.
Need:
- a numeric variable, often with multiple groups
Important considerations:
- box plots summarize data, meaning we can’t see the underlying shape of the distribution or sample size
- add jittered points on top, or if large sample size, consider a violin plot
Box plots are great for a few to multiple groups (too many boxes just results in a lot of information to synthesize, as a viewer). If your x-axis text is long, consider flipping your axes to make them less crunched:
You can modify outlier aesthetics inside geom_boxplot():
Highlight a group of interest – one easy way to do so is by using the {gghighlight} package. Here, we specify a specific month ("October") to highlight:
Since box plots hide sample size, consider overlaying raw data points using geom_jitter() (since our temperature data is too large (>473,000 rows), we’ll use the palmerpenguins::penguins data set to demo):
NOTE: Be sure to remove outliers, since plotting raw data will result in those data points being a second time:
You may have data where you want to include an additional grouping variable – for example, let’s say we want to plot penguin body masses by species and year. We’ll need to at least dodge our overlaid points so that they sit on top of the correct box. Preferably, we both jitter and dodge our points:
Similar to overlaying the raw jittered data points, we can combine our box plot with a beeswarm plot using {ggbeeswarm}. Beeswarm plots visualize the density of data at each point, as well as arrange points that would normally overlap so that they fall next to one another instead. Consider using a standalone beeswarm plot here as well! We’ll again use the palmerpenguins::penguins data set to demo:
What are they?
- Violin plots visualize the distribution of a numeric variable for one or several groups, where the shape of the violin represents the density estimate of the variable (i.e. the more data points in a specific range, the larger the violin is for that range). They provide more information about the underlying distribution than a box plot.
Need:
- a numeric variable, often with multiple groups
Important considerations:
- ordering groups by median value can make it easier to understand
- show sample size when comparing groups with very different distributions (e.g. half violin plot)
Violin plots are great for a few to multiple groups, and are often a better choice than box plots when you have a very large data set (and overlaying jittered points looks busy or downright unreasonable). If your x-axis text is long, consider flipping your axes to make them less crunched:
Overlaying a box plot inside a violin plot can be helpful in providing your audience with summary stats in a compact form:
The {see} package provides geom_violindot(), which is useful for simultaneously visualizing distribution and sample size. Because it can quickly get overcrowded with large sample sizes (like Temp_bot), we’ll use palmerpenguins::penguins to demo here:
