R class - packages and functions
2023-09-01
ggplot2
Create a colored boxplot with the dataset iris, with nice axes titles
Code from the previous class
Use ggplot. The syntax may take a while to get used to, but it is easier to read.
# aes is short for aesthetics, basically what columns hold the values you want to plot
# fill is the color that will be inside the box, color is just for the border
ggplot(data = iris,
aes(x = Species,
y = Sepal.Length,
fill = Species)) +
geom_boxplot() + # the type of plot you want
ylab("Sepal length (mm)") +
xlab("Species epithet")
Improve ggplot by setting a better theme (check here see the options)
# Set a theme for all future plots in this session
theme_set(theme_bw())
# Use different colors to fill and remove the legend
colors = c("red", "green", "blue")
ggplot(data = iris,
aes(x = Species,
y = Sepal.Length,
fill = Species)) +
geom_boxplot(show.legend = F) + # the type of plot you want
ylab("Sepal length (mm)") +
xlab("Species epithet") +
scale_fill_manual(values = colors)
ggplot - part 2
Graphics are layers of data and images put on top of each other. That is why the pieces of the ggplot function are connected by a +.
The codes below construct the plot from the previous slide piece by piece.
labs adds labels:
xandy- will add the labels to the axes (you can use the functionsxlaborylabinstead)title- Normally at the top of the figure (you can use the functionggtitleinstead)subtitle- Goes under the title and is a bit smallercaption- Normally goes at the bottom of the figure and is a small texttag- the letter that goes at the top-left corner of a plot that is one part of several in a figure
dplyr: prepare your data for ggplot
In the example dataset, the names of the groups are in the middle not specified, they are inside the replicate name. We need one column with the values and one with the names of the treatments.
The table now is in the format “wide”, the one we want is called “long”
longData = pivot_longer(
data = rawData,
cols = !Measured, # gets all the columns of the table, except for the one in front of !
names_to = "Replicates", # name of the column that will contain column names from rawData
values_to = "Measurements"
)
str(longData)tibble [48 × 3] (S3: tbl_df/tbl/data.frame)
$ Measured : chr [1:48] "A" "A" "A" "A" ...
$ Replicates : chr [1:48] "Control_1" "Control_2" "Control_3" "Treated_1" ...
$ Measurements: num [1:48] 0.123 0.827 0.149 0.086 0.314 ...Let’s split the values from “Replicates” using str_split from stringr
mutate will return the input table with the new column we create
# pattern is what separates (_)
# i is the part that we want to see: Control_1 has 2 pieces, i=1 returns "Control"
longDataTreatments = longData %>%
mutate(Treatment = str_split_i(Replicates, pattern = "_", i = 1))
str(longDataTreatments)tibble [48 × 4] (S3: tbl_df/tbl/data.frame)
$ Measured : chr [1:48] "A" "A" "A" "A" ...
$ Replicates : chr [1:48] "Control_1" "Control_2" "Control_3" "Treated_1" ...
$ Measurements: num [1:48] 0.123 0.827 0.149 0.086 0.314 ...
$ Treatment : chr [1:48] "Control" "Control" "Control" "Treated" ...
Example violin plot
The code to get a violin plot is the same as the one for a boxplot, the only difference is the geom_violin.
# Set a theme for all future plots in this session
theme_set(theme_bw())
# Use different colors to fill and remove the legend
colors = c("red", "green", "blue")
ggplot(data = iris,
aes(x = Species,
y = Sepal.Length,
fill = Species)) +
geom_violin(show.legend = F) + # violin instead of boxplot
ylab("Sepal length (mm)") +
xlab("Species epithet") +
scale_fill_manual(values = colors)
Types of plots - Bar chart
- Uses
- Visually compare means when the number of replicates is low
- Visually compare counts (number of occurences) - number of students per lab
- How
- Use either
geom_colorgeom_bar(stat = "identity") - Add standard deviation bar with geom_errorbar
geom_errorbar(aes(ymin=Mean-sd, ymax=Mean+sd), width=.2)
- Use either
- identity means that the size of the bar will be the value you put instead of calculating something
- Using either geom_bar or geom_col, you can make a pie chart with
+ coord_polar("y")
df <- data.frame(
group = c("Male", "Female", "Child"),
value = c(25, 25, 50)
)
# Barplot like geom_col
ggplot(df, aes(x = "", y = value, fill = group)) +
# x has to be "" - check how it looks if you put x = group
geom_bar(width = 1, stat = "identity") +
coord_polar("y", start = 1) +
theme_void() # clean up grids, borders, and labels
Types of plot - proportions
You can use geom_bar(stat = "count") if you do not have the number of occurrences of each category computed. This way, ggplot counts it and plots. - Eg. A table with metabolites detected separated by category: how many metabolites of each category were detected?
If you want proportions in %, use stat = "density" instead of stat = "count"
- With geom_bar, you can have the same plot as with geom_col if you use
stat = "identity" - identity means that the size of the bar will be the value you put instead of calculating something
# Datasets about US states
US_statesInfo = data.frame(Name = state.name,
Region = state.region,
Division = state.division)
x = "Number of states"
# Plot the number of states in each division
ggplot(US_statesInfo, aes(y = Division)) +
geom_bar(stat = "count") +
xlab(x)
# Plot the number of states in each region
ggplot(US_statesInfo, aes(y = Region)) +
geom_bar() +
xlab(x)
# Plot the number of states in each division, and color by the region
ggplot(US_statesInfo, aes(y = Division, fill = Region)) +
geom_bar(stat = "count") +
xlab(x)
Types of plots - part 3
Line graph - geom_line
- Visualize data across time - points are connected because they are the same sample at different times.
Let’s say we want to see how the labs of 2 PIs have grown across the years and compare the two.
# Let's create a time series to plot
LabSize = data.frame(Isabel = c(0, 3, 5, 10, 30),
Hugo = c(2, 2, 6, 9, 9),
Year = seq(2014, 2022, 2))
LabSize.Long = pivot_longer(LabSize,
cols = !Year, # all columns from LabSize, except "Year"
names_to = "PI",
values_to = "LabMembers")
ggplot(LabSize.Long,
aes(x = Year, y = LabMembers, color = PI)) +
geom_line()
Advanced line graph
We will use the time series data
'data.frame': 12 obs. of 10 variables:
$ X : chr "Day_1" "Day_2" "Day_3" "Day_4" ...
$ Control_3 : num 0 1.96 3.98 6.18 7.52 ...
$ Control_2 : num 0 2.04 4.02 6 8 ...
$ Control_1 : num 0 2 4 5.82 8.48 ...
$ TreatmentA_1: num 0 2.94 6.03 8.73 12 ...
$ TreatmentA_2: num 0 3 6 9.27 11.28 ...
$ TreatmentA_3: num 0 3.06 5.97 9 12.72 ...
$ TreatmentB_3: num 0 1.53 3 4.5 6.36 ...
$ TreatmentB_2: num 0 1.47 2.98 4.37 5.64 ...
$ TreatmentB_1: num 0 1.5 3.02 4.63 6 ...Let’s transform the data to format long and add a column with the name of the treatment group
library(tidyverse)
timeSeriesLong = pivot_longer(
timeSeries,
cols = !TimePoint,
names_to = c("Treatment", "Replicates"),
names_sep = "_",
values_to = "Growth_measure"
)
head(timeSeriesLong) # A tibble: 6 × 4
TimePoint Treatment Replicates Growth_measure
<dbl> <chr> <chr> <dbl>
1 1 Control 3 0
2 1 Control 2 0
3 1 Control 1 0
4 1 TreatmentA 1 0
5 1 TreatmentA 2 0
6 1 TreatmentA 3 0
Now let’s create another table with the mean, standard deviation and coefficient of variation of each treatment per time point
For the error bars, we cannot simply say “here, this is the standard deviation”.
We need to tell it “use this value as maximum and this as minimum”
See that we set “ymin” and “ymax”. If you have standard deviation for the x-axis, you can add it with “xmin” and “xmax”.
