Pierrette Lo 7/17/2020
- Chapter 7.5-7.8
library(tidyverse)
- Use what you’ve learned to improve the visualisation of the departure times of cancelled vs. non-cancelled flights.
- I copied the below code from the text in section 7.4 and modified it
- Remember you can use
::to call a function from a package without loading the package - Note that performing
mutate()on an existing column (e.g.sched_dep_timebelow) will change that column! I prefer to make a new column with a different name to prevent confusion.
nycflights13::flights %>%
mutate(
cancelled = is.na(dep_time),
sched_hour = sched_dep_time %/% 100,
sched_min = sched_dep_time %% 100,
sched_dep_time = sched_hour + sched_min / 60
) %>%
# ggplot(mapping = aes(sched_dep_time)) +
# geom_freqpoly(mapping = aes(colour = cancelled), binwidth = 1/4)
ggplot(aes(x = cancelled, y = sched_dep_time)) +
geom_boxplot()
This plot makes it more clear that cancelled flights tend to occur later in the day.
As mentioned during our meeting, frequency polygon or density plot would also be a good way to visualize this:
nycflights13::flights %>%
mutate(
cancelled = is.na(dep_time),
sched_hour = sched_dep_time %/% 100,
sched_min = sched_dep_time %% 100,
sched_dep_time = sched_hour + sched_min / 60
) %>%
ggplot(mapping = aes(x=sched_dep_time, y = ..density..)) +
geom_freqpoly(mapping = aes(colour = cancelled), binwidth = .5)
- What variable in the
diamondsdataset is most important for predicting the price of a diamond? How is that variable correlated with cut? Why does the combination of those two relationships lead to lower quality diamonds being more expensive?
I decided to ignore x, y, z, table, and depth since they are
all measurements of size, and I think carat is the most widely used
size-related variable.
First plot the categorical variables (all ordered factors) vs. price.
(You could plot each separately, but I used pivot_longer to get them
all in one facetted plot.) (Also note that pivot_longer only works
with columns of the same data type, so I couldn’t include carat.)
diamonds %>%
select(cut, color, clarity, price) %>%
pivot_longer(cols = -price, names_to = "variable", values_to = "value") %>%
ggplot(aes(x = value, y = price)) +
geom_boxplot() +
facet_wrap(~ variable, scales = "free")
Also plot continuous variable (carat) vs. price:
diamonds %>%
ggplot(aes(x = carat, y = price)) +
geom_point() +
geom_smooth(method = "lm", se = FALSE)## `geom_smooth()` using formula 'y ~ x'
It looks like carat has the strongest relationship with price.
Next, look at carat vs. cut:
diamonds %>%
ggplot(aes(x = cut, y = carat)) +
geom_boxplot()
Looks like the diamonds with the worst cut (Fair) also tend to be the biggest (i.e. most expensive), so that explains why if you just plot cut vs. price, it looks like lower-quality diamonds cost more.
BONUS:
The {patchwork} package is super cool for arranging separate plots. Very useful if you’re trying to make a figure for a paper.
Much more info here, but below is a small example of using {patchwork} instead of my facetted plot above.
library(patchwork)## Warning: package 'patchwork' was built under R version 3.6.2
# Make separate plots and save them as objects
p1 <- diamonds %>%
ggplot(aes(x = carat, y = price)) +
geom_point() +
geom_smooth(method = "lm", se = FALSE)
p2 <- diamonds %>%
ggplot(aes(x = cut, y = price)) +
geom_boxplot()
p3 <- diamonds %>%
ggplot(aes(x = color, y = price)) +
geom_boxplot()
p4 <- diamonds %>%
ggplot(aes(x = clarity, y = price)) +
geom_boxplot()
# Then arrange them using {patchwork} notation
p1 + (p2 / p3 / p4)## `geom_smooth()` using formula 'y ~ x'
- Install the ggstance package, and create a horizontal boxplot. How does this compare to using coord_flip()?
In older versions of {ggplot2}, the categorical variable for a bar or
box plot always needed to be on the x axis, and you would need to use
coord_flip() if you wanted to make the plot horizontal. However, in
these plots the vertical axis is still considered to be the x axis,
which could get confusing.
A recent update to {ggplot2} has removed this requirement, so you can just plot the categorical variable on the y axis - no need for mental gymnastics or installing a separate package.
# diamonds %>%
# ggplot(aes(x = cut)) +
# geom_bar() +
# coord_flip()
diamonds %>%
ggplot(aes(y = cut)) +
geom_bar()
- One problem with boxplots is that they were developed in an era of much smaller datasets and tend to display a prohibitively large number of “outlying values”. One approach to remedy this problem is the letter value plot. Install the lvplot package, and try using geom_lv() to display the distribution of price vs cut. What do you learn? How do you interpret the plots?
This package isn’t on CRAN, so you’ll need to install from GitHub (instructions here). You’ll also need to have package {devtools} installed first.
The statistical methods are described in the help (?geom_lv).
It was pointed out during the meeting that the widest box in geom_lv
corresponds to the IQR, which you can see by superimposing a boxplot.
# install.packages("devtools")
# devtools::install_github("hadley/lvplot")
library(lvplot)
diamonds %>%
ggplot(aes(x = cut, y = price)) +
geom_lv(fill = "orange") +
geom_boxplot(alpha = 0.5)
- Compare and contrast geom_violin() with a facetted geom_histogram(), or a coloured geom_freqpoly(). What are the pros and cons of each method?
Freqpoly makes it easier to compare different cuts for a specific price.
However it’s a bit difficult to interpret the overall differences. Using
y = ..density.. helps a bit as it gives them the same scale
(proportions of 1).
diamonds %>%
ggplot(aes(x = price, y = ..density.., color = cut)) +
geom_freqpoly()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
A violin plot is like a boxplot that shows density. Violin plots and facetted histograms are better than freqpoly for comparing overall distributions, but worse for trying to compare a specific price point.
diamonds %>%
ggplot(aes(x = cut, y = price)) +
geom_violin()
diamonds %>%
ggplot(aes(x = price, y = ..density..)) +
geom_histogram() +
facet_wrap(~ cut)## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
One of our study group participants mentioned that his favorite plot type is a dotplot. You can see the distribution and accurately read the y-axis without the randomness introduced by jittering.
ggplot(mpg,aes(class, hwy)) +
geom_dotplot(binwidth = .3, binaxis='y', stackdir='center',
stackratio=1.5, dotsize=1.2)
- If you have a small dataset, it’s sometimes useful to use geom_jitter() to see the relationship between a continuous and categorical variable. The ggbeeswarm package provides a number of methods similar to geom_jitter(). List them and briefly describe what each one does.
Beeswarm plot is like a jittered scatterplot + violin plot.
Sample plots using {ggbeeswarm} can be found here
It’s hard to see anything on these plots when the dataset is too large,
so I used sample_n to get a random sample (could also use
sample_frac):
# install.packages("ggbeeswarm")
library(ggbeeswarm)## Warning: package 'ggbeeswarm' was built under R version 3.6.3
diamonds %>%
sample_n(size = 1000) %>%
ggplot(aes(x = cut, y = price)) +
geom_quasirandom(alpha = 0.5)
Unrelated to beeswarm plots -
With a small-ish dataset, I like to use boxplot (with outliers hidden) + scatterplot. I think the boxplot is more recognizable to scientists and having the points gives more information about the distribution.
I also like to make the box width proportional to number of observations in each category.
diamonds %>%
sample_n(size = 500) %>%
ggplot(aes(x = cut, y = price)) +
geom_boxplot(outlier.shape = NA, varwidth = TRUE, color = "darkred") +
geom_jitter(alpha = 0.5, width = 0.2)