In graeberlab-ucla/glab.library: Data Analysis Functions

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(glab.library)
library(glab.library)
library(ggplot2)

head(iris)

1. The Basics

Set up mapping aesthetics by mapping continuous variables (columns in our dataset) to x and y axes.

ggplot(data = iris, mapping = aes(x = Sepal.Length, y = Sepal.Width))

Creating a scatterplot by adding a geom_point layer

ggplot(iris, aes(x = Sepal.Length, y = Sepal.Width)) +
geom_point()

Add color by species of plant to the aesthetics mapping

ggplot(iris, aes(x = Sepal.Length, y = Sepal.Width, color = Species)) +
geom_point()

Adding multiple layers

Set up mapping aesthetics with discrete x variable and continuous y variable.

ggplot(iris, aes(x = Species, y = Sepal.Width, color = Species)) +
geom_point()

Try a different geometric layer boxplot.

ggplot(iris, aes(x = Species, y = Sepal.Width, color = Species)) +
geom_boxplot()

or a violin plot.

ggplot(iris, aes(x = Species, y = Sepal.Width, color = Species)) +
geom_violin()

Add second layer for points.

ggplot(iris, aes(x = Species, y = Sepal.Width, color = Species)) +
geom_violin() + geom_point()

Note that the order of the layers matters! The first ones are plotted first and new layers are added on top (Only the violins are visible)

ggplot(iris, aes(x = Species, y = Sepal.Width, color = Species)) +
geom_point() + geom_violin()

geom_jitter for points with some random noise added.

If the points are overlapping each other, try this.

ggplot(iris, aes(x = Species, y = Sepal.Width, color = Species)) +
geom_violin() + geom_jitter(width = 0.33, height = 0.05)

Beeswarm Plots: More organized points with random noise

Jittered points help but some points may still be overlapping, so let's use Beeswarm Plots.

library(ggbeeswarm)
ggplot(iris, aes(x = Species, y = Sepal.Width, color = Species)) +
geom_violin() + geom_beeswarm(size = 1.5, cex = 1.5)

Mapping by layer

Color one specific layer of ggplot

Suppose we only want one layer to be colored, rather than all of them. Here we move mapping of color to a specific layer of the ggplot.

ggplot(iris, aes(x = Species, y = Sepal.Width)) +
geom_violin(aes(color = Species)) + geom_beeswarm(size = 1.5, cex = 1.5)

Facets

Mapping continuous variables to x and y axes.

Set up mapping aesthetics by mapping continuous variables to x and y axes and coloring by "Treatment".

head(CO2)
gg <- ggplot(CO2, aes(x = conc, y = uptake, color = Treatment)) +
geom_point()
print(gg)

Facet Wrap

Our data comes from two different sites. Let's separate the plots by the collection site using facet_wrap.

gg2 <- gg + facet_wrap(~ Type)
print(gg2)

Adding a smooth fitted line to help visualize the trend.

gg2 + geom_smooth()

What if we don't want the lines colored?

We can try removing the color mapping in favor of a constant mapping to "black". However, this will cause an issue! The color aesthetic distinguished the two lines we want to draw per facet.

gg2 + geom_smooth(color = "black")

Use the group aesthetic mapping when you want to separate mappings without color/fill/etc.

gg2 + geom_smooth(aes(group = Treatment), color = "black", se = F)

Method Parameter

Although it does not make sense for this data in particular, we can change the method parameter to define what kind of line is fit to the data.

gg2 + geom_smooth(aes(group = Treatment), color = "black", se = F, method = "lm")

Barcharts and Stacking

head(mtcars)

Set up mappings

gg <- ggplot(mtcars, aes(x = factor(gear), fill = factor(carb)))

Stacked barchart of counts, colored by the number of carburetors.

gg + geom_bar(position = "stack")

Without the fill parameter, the plot will show the count of each gear.

ggplot(mtcars, aes(x = factor(gear))) + geom_bar()

Stacked barchart with relative proportions

gg + geom_bar(position = "fill")

Dodged barchart with counts

gg + geom_bar(position = "dodge")

Integrating dpylr

If we want to plot the relative frequencies AND still dodge the fill mapping we will need to use dplyr to calculate the values ourselves instead of leaving it up to R.

library(dplyr)

Calculate relative frequencies using dplyr piping

mtcars %>%
group_by(gear, carb) %>%
summarize(n = n()) %>%
mutate(freq = n / sum(n)) %>%
ggplot(aes(x = factor(gear), y = freq, fill = factor(carb))) +
geom_bar(stat = "identity", position = "dodge")

2. Detailing and Saving Plots

gg <- ggplot(iris, aes(x = Sepal.Length, y = Sepal.Width, color = Species)) +
geom_point() + stat_ellipse(show.legend = F)
print(gg)

gg + theme_classic()
gg + theme_minimal()
gg + theme_bw()

Themes

Some other themes I like to use are theme_bw() and theme_minimal() link to list of themes: https://ggplot2.tidyverse.org/reference/ggtheme.html

library(RColorBrewer)

Incorporating many elements

preset theme + custom theme (custom elements must come after preset!!) setting axis limits with xlim/ylim changing labels changing colors with RColorBrewer

pretty_gg <- gg +
theme_classic() +
theme(axis.text = element_text(size = 9, color = "black"),
legend.position = "bottom", plot.title = element_text(hjust=0.5, size = 10),
axis.title = element_text(size = 9)) +
ylim(2, 4.25) +
labs(x = "Sepal Length", y = "Sepal Width", fill = "Species",
title = "Sepal Length vs. Sepal Width across Iris Species") +
scale_color_brewer(palette = "Dark2")
print(pretty_gg)

Saving Images

Save image as .png using "cairo" graphics device ("cairo" for windows, "Xlib" or "quartz" for mac) you may need to install the Cairo package with install.packages("Cairo") [this may also work for mac]

png(filename = "Iris Sepal Length vs Sepal Width.png", res = 300, type = "Xlib",
    height = 1200, width = 1200)
print(pretty_gg)
dev.off()

Without setting type to cairo, the resulting image quaility will be lower.

This particularly affects curved lines and circles

png(filename = "Iris Sepal Length vs Sepal Width no Cairo.png", res = 300, #won't be as clear
    height = 1200, width = 1200)
print(pretty_gg)
dev.off()

Set the units to inches for ease of use.

png(filename = "Iris Sepal Length vs Sepal Width.png", res = 300, type = "cairo",
    units = "in", height = 4, width = 4)
print(pretty_gg)
dev.off()

Another way to save images is with the ggpubr package.

library(ggpubr)
ggsave("Iris Sepal Length vs Sepal Width.png", pretty_gg, dpi = 300, type = "cairo",
    height = 4, width = 4)

Ordering factors in boxplot/barchart

Simple boxplot

ggplot(CO2, aes(x = Type, y = uptake, fill = Treatment)) + geom_boxplot()

Create new factor combining Treatment and Type.

This allows us to use more colors if needed.

CO2$interaction <-  factor(paste0(CO2$Treatment, CO2$Type),
    levels = c("nonchilledMississippi", "chilledMississippi",
    "nonchilledQuebec", "chilledQuebec"))

Customized ggplot boxplot

1. Setting x axis order of factors
2. Manually changing fill palette, still using RColorBrewer brewer.pal()
3. Changing labels
4. Preset theme + custom theme

boxplot_gg <- ggplot(CO2, aes(x = Type, y = uptake, fill = interaction)) +
geom_boxplot(color = "black", outlier.size = 0.5) +
scale_x_discrete(limits = c("Mississippi", "Quebec")) +
scale_fill_manual(values = brewer.pal("Paired", n = 9)[c(1,2,5,6)],
    labels = c("control (M)", "chilled (M)", "control (Q)", "chilled (Q)")) +
labs(x = "", y = expression(CO[2]~Uptake~(mu*mol/m^2~sec)), fill = "Treatment") +
theme_classic() +
theme(axis.text.x = element_text(size = 9, color = "black", angle = 45, hjust = 1),
    axis.text.y = element_text(size = 8, color = "black"))

Works the same as print(boxplot_gg) if you have a graphics device open.

boxplot_gg
ggsave("CO2 Uptake by Location by Treatment.png", boxplot_gg, dpi = 300, type = "cairo",
    height = 3, width = 3)

3. Wide vs. Long Data Format

head(USPersonalExpenditure)

Cast as data frame for compatiblity with ggplot.

R doesn't like numeric column names, and adds an X to the beginning.

df <- data.frame(USPersonalExpenditure)

Use melt to transform the data into long format.

library(reshape)
USPE_melt <- melt(USPersonalExpenditure)

Add some descriptive column names.

colnames(USPE_melt) <- c("variable", "year", "value")

Scatterplot comparing expenditures in 1940 and 1945 with wide format data.

Here, coord_fixed() forces the axes to the same scale. geom_label adds text labels to the plot, but they are cutoff by the plot margins.

ggplot(df, aes(x = X1940, y = X1945)) +
geom_point() +
geom_abline() +
geom_label(aes(label = rownames(df))) +
coord_fixed() +
labs(x = "Expenditures in 1940 (Billions of Dollars)",
    y = "Expenditures in 1945 (Billions of Dollars)")

We can use geom_label_repel (and geom_text_repel) to add non-overlapping labels.

THIS ONE IS MUCH BETTER. NO OVERLAP.

library(ggrepel)
ggplot(df, aes(x = X1940, y = X1945)) +
geom_point() +
geom_abline() +
geom_label_repel(aes(label = rownames(df)), force = 20) +
coord_fixed() +
labs(x = "Expenditures in 1940 (Billions of Dollars)",
    y = "Expenditures in 1945 (Billions of Dollars)")

Another way to visualize the difference in expenditures with long format data.

USPE_melt %>%
filter(year %in% c(1940, 1945)) %>%
ggplot(aes(x = factor(year), y = value, group = 1)) +
geom_line() +
geom_point() +
facet_wrap(~ variable) +
labs(x = "Year", y = "Total US Expenditures (Billions of Dollars)") +
theme_bw()

If we scale the data independently it may become less intuitive.

USPE_melt %>%
filter(year %in% c(1940, 1945)) %>%
ggplot(aes(x = factor(year), y = value, group = 1)) +
geom_line() +
geom_point() +
facet_wrap(~ variable, scale = "free_y") +
labs(x = "Year", y = "Total US Expenditures (Billions of Dollars)") +
theme_bw()

Long format data enables us to look at even more data points.

ggplot(USPE_melt, aes(x = factor(year), y = value, group = 1)) +
geom_line() +
geom_point() +
facet_wrap(~ variable) +
labs(x = "Year", y = "Total US Expenditures (Billions of Dollars)") +
theme_bw()

Specifying certain columns you do not want in long format.

These wil be for factor levels you plan on using later while plotting.

mtcars_melt <- melt(mtcars, id.vars = c("vs", "am", "gear", "carb"))

Another example combining many layers and parameters.

ggplot(mtcars_melt, aes(y = value, x = factor(am))) +
geom_violin() +
geom_beeswarm(aes(color = factor(vs)), cex = 5) +
facet_wrap(~ variable, scale = "free_y") +
theme_bw() +
scale_x_discrete(limits = factor(c(0,1)), labels = c("Automatic", "Manual")) +
labs(x = "", y = "Value", color = "Engine Type", title = "Automatic\nvs.\nManual\nMotor Vehicle Trends") +
scale_color_manual(values = c("black", "red"), labels = c("V-shaped", "Straight"))

Here, independent scaling of the y-axis is preferred.

If we don't some variables can't be viewed properly.

ggplot(mtcars_melt, aes(y = value, x = factor(am))) +
geom_violin() +
geom_beeswarm(aes(color = factor(vs)), cex = 5) +
facet_wrap(~ variable) +
theme_bw() +
scale_x_discrete(limits = factor(c(0,1)), labels = c("Automatic", "Manual")) +
labs(x = "", y = "Value", color = "Engine Type", title = "Automatic\nvs.\nManual\nMotor Vehicle Trends") +
scale_color_manual(values = c("black", "red"), labels = c("V-shaped", "Straight"))

graeberlab-ucla/glab.library documentation built on Oct. 28, 2024, 10:48 a.m.