Getting started

Clear the environment
Load the following packages

library(knitr)
library(dplyr)
library(stringr)
library(reshape2)
library(magrittr)

We will use 2 built-in datasets for data visualisation

A set of samples with measurements of several traits

• iris contains petal and sepal measurements for irises

A time series

• ChickWeight contains weights for chicks on different diets

Load and configure the dataset iris

data(iris)

iris %<>% as_data_frame() # converts it to a newer data frame format

iris

R has plotting functions in the base package graphics

?plot()
?boxplot()
?hist()

And other plotting packages have been written

We will focus on the function ggplot()

ggplot2 is flexible and powerful

Has a unique approach and syntax

Consists of layers of plotting information

The main function in ggplot2 is ggplot()

library(ggplot2)

4 basic steps to plotting with ggplot():

• Choose the dataset to use
  
• Choose the data to plot using aes() = aesthetics
  
• Choose the way to display the data with a geometry
  
• Then make it pretty!

The first value is the dataset to use

ggplot(data = iris)

is the same as

ggplot(iris)

We have a blank canvas ie the plot area

aes() defines what is plotted, colour and shape etc

ggplot(iris, aes(x = Species, y = Petal.Length))

Now we have a grid and axes

Note that we have chosen to plot
- a numeric variable (Petal.Length) vs
- a categorical variable (Species)

This will constrain what geometries we can use

Add a geometry using various geom_...() functions

ggplot(iris, aes(x = Species, y = Petal.Length)) +    # note the +
    geom_point()

Colour the points by species

ggplot(iris, aes(x = Species, y = Petal.Length, colour = Species)) +
    geom_point()

We can modify the points in many ways eg

• size()
  
• shape()
  
• alpha() - transparency
  
• etc

Add a main title
Clean up the y axis title
Remove the figure legend

ggplot(iris, aes(x = Species, y = Petal.Length, colour = Species)) +
    geom_point() +
    labs(title = "Iris Petal Length by Species", y = "Petal Length") +
    theme(plot.title = element_text(hjust = 0.5), 
          legend.position = "none")

There are many options under theme()

The points are tightly clustered on the species line

We can stagger them using geom_jitter()

ggplot(iris, aes(x = Species, y = Petal.Length, colour = Species)) +
    geom_jitter() +
    labs(title = "Iris Petal Length by Species", y = "Petal Length") +
    theme(plot.title = element_text(hjust = 0.5), 
          legend.position = "none")

Use geom_boxplot() to display the data

ggplot(iris, aes(x = Species, y = Petal.Length, colour = Species)) +
    geom_boxplot() +
    labs(title = "Iris Petal Length by Species", y = "Petal Length") +
    theme(plot.title = element_text(hjust = 0.5), 
          legend.position = "none")

We can plot more than one geometry onto the same plot
- geom_boxplot()
- geom_jitter()

ggplot(iris, aes(x = Species, y = Petal.Length, colour = Species)) +
    geom_boxplot(colour = "darkgrey", alpha = 0) +
    geom_jitter(alpha = 0.5) +
    labs(title = "Iris Petal Length by Species", y = "Petal Length") +
    theme(plot.title = element_text(hjust = 0.5), 
          legend.position = "none")

Embed aes() within each geom()

ggplot(iris) +
    geom_boxplot(aes(x = Species, y = Petal.Length), 
                 colour = "darkblue", alpha = 0) +
    geom_boxplot(aes(x = Species, y = Petal.Width), 
                 colour = "darkgreen", alpha = 0) +
    geom_boxplot(aes(x = Species, y = Sepal.Width), 
                 colour = "purple", alpha = 0) +
    geom_boxplot(aes(x = Species, y = Sepal.Length), 
                 colour = "darkred", alpha = 0) +
    labs(title = "Measurements by Iris Species", y = "Size") +
    theme(plot.title = element_text(hjust = 0.5))

We will now plot
- a numeric variable (Petal.Length) vs
- another numeric variable (Species)

Again this will constrain what geometries we can use

Start with geom_point()

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

ggplot(iris, aes(x = Sepal.Length, y = Petal.Length, 
                 colour = Species)) +
    geom_point() +
    labs(title = "Iris Petal Length vs Sepal Length", 
         x = "Sepal Length", y = "Petal Length") +
    theme(plot.title = element_text(hjust = 0.5, face = "bold"))

Use geom_smooth()
Add line for each species

ggplot(iris, aes(x = Sepal.Length, y = Petal.Length, 
                        colour = Species)) +
    geom_point() +
    geom_smooth(method = "lm") +
    labs(title = "Iris Petal Length vs Sepal Length", 
         x = "Sepal Length", y = "Petal Length") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"))

Correlation between sepal and petal length

library(corrplot)

cor(iris$Sepal.Length, iris$Petal.Length, 
    use = "pairwise.complete.obs")

Correlation between all 4 numerical variables

iris
correlations <- cor(iris %>% select(-Species), 
                    use = "pairwise.complete.obs")
correlations

corrplot(correlations, method = "shade", shade.col = NA, 
         tl.col = "black", tl.srt = 90, order = "AOE", 
         mar=c(2,0,2,0))

With values

corrplot(correlations, method = "shade", shade.col = NA, 
         tl.col = "black", tl.srt = 90, order = "AOE", 
         mar=c(2,0,2,0), addCoef.col = "darkgrey")

Load and configure the dataset ChickWeight
And fix column names!

data(ChickWeight)

ChickWeight %<>% as_data_frame() # newer data frame format
ChickWeight

colnames(ChickWeight) <- c("Weight", "Day", "Chick", "Diet") 
ChickWeight

DietCount <- ChickWeight %>%    # take starting df
    select(Diet, Chick) %>%     # keep 2 of the columns
    unique %>%                  # keep 1 copy of each combination
    count(Diet) %>%             # count number of chicks by diet
    rename(Count = n)           # rename the count column
DietCount

Use geom_col()

ggplot(DietCount, aes(x = Diet, y = Count, fill = Diet)) + 
    geom_col() +
    labs(title = "Number of chicks on each diet") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
          legend.position = "none")

And again with diets as integers - this is cheating!

ggplot(DietCount, aes(x = Diet, y = Count, fill = as.integer(Diet))) +
    geom_col() +
    labs(title = "Number of chicks on each diet") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
          legend.position = "none")

Use geom_histogram() to get the distributions
Use facet_wrap() to plot each day in its own plot

ggplot(ChickWeight, aes(x = Weight)) +
    geom_histogram(fill = "darkblue") +
    labs(title = "Distribution of Chick Weight by Day", y = "Count") +
    facet_wrap(~ Day, scales = "free") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
          axis.text.x = element_text(angle = 45, hjust = 1), 
          legend.position = "none")

Back to geom_point() to create a scatter plot

ggplot(ChickWeight, aes(x = Day, y = Weight, colour = Chick)) +
    geom_point() +
        labs(title = "Chick Weights by Day") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
          legend.position = "none")

Use to geom_line() instead of geom_point()

ggplot(ChickWeight, aes(x = Day, y = Weight, colour = Chick)) +
    geom_line() +
    labs(title = "Chick Weights by Day") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
          axis.text.x = element_text(angle = 45, hjust = 1), 
          legend.position = "none")

Group by chick so the data from day to day remains connected But colour by diet

ggplot(ChickWeight, 
       aes(x = Day, y = Weight, group = Chick, colour = Diet)) +
    geom_line() +
    labs(title = "Chick Weights by Day") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"))

Want mean, SD and 95% CI to compare the diets

WeightSummary <- ChickWeight %>%
    group_by(Diet, Day) %>%
    summarise(N = length(Weight) %>% as.integer(), 
              Mean = mean(Weight) %>% round(digits = 1), 
              SD = sd(Weight) %>% round(digits = 1)) %>%
    mutate(MeanSD = paste(Mean, SD, sep = " ± ")) %>%
    mutate(CI95 = (qt(0.975,df = N-1)*SD/sqrt(N)) %>% 
               round(digits = 1)) %>%
    mutate(MeanCI95 = paste(Mean, CI95, sep = " ± "))
WeightSummary

ggplot(WeightSummary, aes(x = Day, y = Mean, group = Diet, 
                          colour = Diet)) +
    geom_line() +
    labs(title = "Chick Weights over Day") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
          axis.text.x = element_text(angle = 45, hjust = 1))

The diets are not in order of increasing effectiveness
Reorder them

WeightSummary$Diet
WeightSummary$Diet <- ordered(WeightSummary$Diet, 
                              levels = c("1", "2", "4", "3"))
WeightSummary$Diet

ggplot(WeightSummary, aes(x = Day, y = Mean, group = Diet, 
                          colour = Diet)) +
    geom_line() +
    labs(title = "Chick Weights over Day") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
          axis.text.x = element_text(angle = 45, hjust = 1))

Use geom_ribbon()

ggplot(WeightSummary, aes(x = Day, y = Mean, colour = Diet)) +
    geom_ribbon(aes(ymin = Mean-CI95, ymax = Mean+CI95, 
                    color = Diet), 
                alpha = 0) +
    labs(title = "Chick Weights over Time") +
    theme(plot.title = element_text(hjust = 0.5, face="bold"), 
    axis.text.x = element_text(angle = 45, hjust = 1))

Go back to the data
Is your visual interpretation consistent?

Convert the data from long to wide format

WeightSummary2 <- WeightSummary %>% 
    dcast(Diet ~ Day, value.var = "MeanCI95")

WeightSummary2 %>% dim()
WeightSummary2[,c(1,13)]