ggplot2 is a hugely powerful R package which enables the creation of high quality scientific charting and visualisation for communication.
This note walks through some basics for creating charts to illustrate the literature review but there are lots of resources for chart creation of which this course really showcases what is possible.
Load libraries and import data
Code
## load librarieslibrary(pacman)pacman::p_load(tidyverse, readxl, here, skimr, overviewR, ggmap, gt, gtsummary, DataExplorer, gtExtras, rphylopic, magick) ## readxl is needed to load excel files into Rpath <- here::here("data") ## the `here` package helps with file pathsxls <-list.files(path, "xls", full.names =TRUE)data <- readxl::read_xlsx(xls[1]) |> janitor::clean_names()
The dataset is wide - that is each variable or category has its own column, but for analysis and plotting it might be easier to groups some of the variables to make the data more hierarchical. For example climate change outcome has a set of outcome categories each of which have a direction of change.
The skimr package gives us a rapid overview of the data structure…
## Select categorical variableschr_vars <-select_if(data, is.character) |> janitor::clean_names() ## this function converts variable names to snake case which makes them much more consistentglimpse(chr_vars)
We’ll use ggplot2 which is a hugely powerful and (hopefully) intuitive package for high quality production charting.
Lets plot climate_change_outcome
The first step is to count the frequency of each category within the variable:
Code
## the |> symbol is called a "pipe" . It says "do the instruction on the right hand side to the data on the left hand side. So... count categories in the variable climate_effect_outcome in the data set chr_varscounts <- chr_vars |>count(climate_effect_outcome)counts
We can now construct a bar (column) chart using gg plot
Code
counts |>ggplot(aes(x = climate_effect_outcome, y = n)) +## we have to wrap x and y variables in aes()geom_col() ## geom_col() plots bars
It’s a bit of a mess…
Lets do 2 things:
Order the bars by size
Flip the chart from horizontal to vertical
Note that ggplot is built on the idea that charts are built in layers - axes, annotations, data, panels and plot areas and so on, and each layer we add needs a + sign -
In the code below we tell ggplot to add columns to represent the data on the chart by + geom_col()
That’s better but we can see a few issues:
Labels are ?duplicated and rather messy
The y axis label is redundant
The x axis needs to be integer and we need to label it correctly
We need to add titles etc
We should remove (or relabel) NAs.
Let’s tackle these.
Labelling or modifying axes labels, adding titles
This is controlled by labs()
Code
g <- counts |>ggplot(aes(reorder(climate_effect_outcome, n), n)) +geom_col(fill ="blue") +## (fill = ... controls the colour of the bars)labs(y ="No of studies", x =NULL, title ="Climate outcomes") +coord_flip() g
Lets move the title to the left, and reduce the font size of the y axis labels, and we’ll change the scale of the x-axis
Positions of titles, axes labels, font sizes and so on are controlled by theme()
Code
g <- g +theme(plot.title.position ="plot", axis.text.x =element_text(size = .5))g
Further modifications
The value labels still need tidying but that may best be done in the data. We can change the colour of the bars and the background.
Code
g +theme(panel.background =element_blank()) ## removes panel background
Putting it all together
Code
chr_vars |>count(climate_effect_outcome) |>ggplot(aes(reorder(climate_effect_outcome, n), n)) +geom_col(fill ="blue") +labs(y ="No of studies", x =NULL, title ="Climate outcomes") +ylim(c(0, 14)) +coord_flip() +theme(plot.title.position ="plot", axis.text.y =element_text(size =7), panel.background =element_blank())
Making it generic - writing a function
Now we have a basic template, it would be useful to reuse this for other variables.
To do this we can write a function - like a macro - which in R is very easy - just need to wrap our code in function() and identify the input we want to change - which in this case is other variables. Lets call the function plot_ordered_bar_chart. The core looks like this…
Code
plot_ordered_bar_chart <-function(df, var){ df = df var <-enquo(var) df |>count(!!var) |>ggplot(aes(reorder(!!var, n), n)) +geom_col(fill ="yellow") +labs(y ="No of studies", x =NULL, title = var) +ylim(c(0, 14)) +coord_flip() +theme(plot.title.position ="plot", axis.text.y =element_text(size =7), panel.background =element_blank())}plot_ordered_bar_chart(chr_vars, var= climate_effect_outcome)
We might want to compare frequencies across multiple variables, for example
Code
## lets create a frequency table of the number of breeds mentioned in studes for each herbivore groupgrouped <- data |>select(contains("vine")) |>pivot_longer(names_to ="herbivore", values_to ="breed", cols =1:6) |>group_by(herbivore) |>count(breed)grouped
---title: "Visualising herbivory and climate change literature"format: htmleditor: visualexecute: cache: truetoc: truetoc-location: leftcode-fold: truecode-tools: true---`ggplot2` is a hugely powerful R package which enables the creation of high quality scientific charting and visualisation for communication.This note walks through some basics for creating charts to illustrate the literature review but there are lots of resources for chart creation of which [this course](https://rstudio-conf-2022.github.io/ggplot2-graphic-design/) really showcases what is possible.## Load libraries and import data```{r}#| label: import data ## load librarieslibrary(pacman)pacman::p_load(tidyverse, readxl, here, skimr, overviewR, ggmap, gt, gtsummary, DataExplorer, gtExtras, rphylopic, magick) ## readxl is needed to load excel files into Rpath <- here::here("data") ## the `here` package helps with file pathsxls <-list.files(path, "xls", full.names =TRUE)data <- readxl::read_xlsx(xls[1]) |> janitor::clean_names()``````{r}introduce(data)```## Data wranglingThe dataset is wide - that is each variable or category has its own column, but for analysis and plotting it might be easier to groups some of the variables to make the data more hierarchical. For example *climate change outcome* has a set of outcome categories each of which have a direction of change.The `skimr` package gives us a rapid overview of the data structure...```{r}skimr::skim(data)```Lets try and reshape the climate effects variables.```{r}#| label: select climate effect variablesclimate_effect <- data |>select(item, title, albedo:direction_of_climate_change_outcome)climate_effect_long <- climate_effect |>pivot_longer(names_to ="vars", values_to ="vals", cols =3:16)pluck(climate_effect_long, "vars") |>unique()```### Recoding variables to create groups```{r}cel <- climate_effect_long |>mutate(var_cat =case_when(str_detect(vars, "methane") ~"methane", str_detect(vars, "biomass") ~"biomass", str_detect(vars, "emissions") ~"emissions",TRUE~ vars)) gt <- cel |> janitor::tabyl(vars, vals) |> gt::gt() gt |>tab_header(title ="Climate change outcomes") |>tab_spanner(label ="Outcome",columns =2:9 ) |>tab_row_group(label ="Biomass", rows =c(1, 11) ) |>tab_row_group(label ="Methane", rows =c(5:9) ) ```Lets do similar data on herbivores.```{r}#| label: download phylopic imagesdir.create(paste0(here(), "/images"))image_path <-here("images")taxa <-"sheep"ns <-name_search(taxa, options ="namebankID")[1,]id <-name_images(uuid = ns$uid[1])uid <- id$same[[4]]$uidsize <-64img <-paste0("http://phylopic.org/assets/images/submissions/", uid, ".", size, ".png") %>%image_read()img |>image_write(paste0(image_path, "/sheep.png"))```### Table decoration with images```{r}#| label: add images to table## Create a table of imagesimage_files <-list.files(here::here("images"), "png", full.names = T) taxa <-c("Cows", "Goat", "Horses", "Sheep")image_tab <-tibble(taxa, image_files)data |>select(item, contains("ine"), -continent_and_country) |>pivot_longer(names_to ="Herbivores", values_to ="Breed", cols =2:11, values_drop_na =TRUE) |>mutate(Herbivore_Group =case_when(str_detect(Herbivores, "ines") ~"Outcome", TRUE~"Comparator")) |>group_by(Herbivore_Group, Herbivores) |>count(Breed) |>mutate(Herbivores =str_remove_all(Herbivores, "_.*ine?."), Herbivores =str_to_title(Herbivores)) |>pivot_wider(names_from ="Herbivore_Group", values_from ="n", values_fill =0) |>mutate(Breed =str_remove(Breed, "-"), Breed =str_remove(Breed, "\\r\\n")) |>left_join(image_tab, by =c("Herbivores"="taxa")) |>select(Herbivory = image_files, Breed, Outcome, Comparator) |> gt::gt() |>gt_img_rows(Herbivory, img_source ="local") |>gt_theme_guardian()```## Basic template for plotting categorical variables### Select categorical variables```{r}## Select categorical variableschr_vars <-select_if(data, is.character) |> janitor::clean_names() ## this function converts variable names to snake case which makes them much more consistentglimpse(chr_vars) ```### Ordered bar charts (recommended)We'll use `ggplot2` which is a hugely powerful and (hopefully) intuitive package for high quality production charting.Lets plot `climate_change_outcome`The first step is to count the frequency of each category within the variable:```{r}## the |> symbol is called a "pipe" . It says "do the instruction on the right hand side to the data on the left hand side. So... count categories in the variable climate_effect_outcome in the data set chr_varscounts <- chr_vars |>count(climate_effect_outcome)counts```We can see there are `r nrow(counts)` in the data.We can now construct a bar (column) chart using gg plot```{r}counts |>ggplot(aes(x = climate_effect_outcome, y = n)) +## we have to wrap x and y variables in aes()geom_col() ## geom_col() plots bars ```It's a bit of a mess...Lets do 2 things:1. Order the bars by size2. Flip the chart from horizontal to verticalNote that `ggplot` is built on the idea that charts are built in layers - axes, annotations, data, panels and plot areas and so on, and each layer we add needs a `+` sign -In the code below we tell ggplot to add columns to represent the data on the chart by `+ geom_col()`That's better but we can see a few issues:1. Labels are ?duplicated and rather messy2. The y axis label is redundant3. The x axis needs to be integer and we need to label it correctly4. We need to add titles etc5. We should remove (or relabel) NAs.Let's tackle these.### Labelling or modifying axes labels, adding titlesThis is controlled by `labs()````{r}g <- counts |>ggplot(aes(reorder(climate_effect_outcome, n), n)) +geom_col(fill ="blue") +## (fill = ... controls the colour of the bars)labs(y ="No of studies", x =NULL, title ="Climate outcomes") +coord_flip() g```### Lets move the title to the left, and reduce the font size of the y axis labels, and we'll change the scale of the x-axisPositions of titles, axes labels, font sizes and so on are controlled by `theme()````{r}#| label: labellingg <- g +theme(plot.title.position ="plot", axis.text.x =element_text(size = .5))g```### Further modificationsThe value labels still need tidying but that may best be done in the data. We can change the colour of the bars and the background.```{r}g +theme(panel.background =element_blank()) ## removes panel background```### Putting it all together```{r}#| label: putting it all togetherchr_vars |>count(climate_effect_outcome) |>ggplot(aes(reorder(climate_effect_outcome, n), n)) +geom_col(fill ="blue") +labs(y ="No of studies", x =NULL, title ="Climate outcomes") +ylim(c(0, 14)) +coord_flip() +theme(plot.title.position ="plot", axis.text.y =element_text(size =7), panel.background =element_blank())```### Making it generic - writing a functionNow we have a basic template, it would be useful to reuse this for other variables.To do this we can write a function - like a macro - which in R is very easy - just need to wrap our code in `function()` and identify the input we want to change - which in this case is other variables. Lets call the function `plot_ordered_bar_chart`. The core looks like this...```{r}#| label: plotting functionplot_ordered_bar_chart <-function(df, var){ df = df var <-enquo(var) df |>count(!!var) |>ggplot(aes(reorder(!!var, n), n)) +geom_col(fill ="yellow") +labs(y ="No of studies", x =NULL, title = var) +ylim(c(0, 14)) +coord_flip() +theme(plot.title.position ="plot", axis.text.y =element_text(size =7), panel.background =element_blank())}plot_ordered_bar_chart(chr_vars, var= climate_effect_outcome)``````{r}#| eval: falsevariables <-colnames(chr_vars)plot_ordered_bar_chart(chr_vars, horses_equines)```### Pie chart (not recommendedPie charts are unexpectedly hard to to create in R - there is no geom in ggplot for example```{r}#| eval: truecounts |>ggplot(aes(x="", y=n, fill=climate_effect_outcome)) +geom_bar(stat="identity", width=1) +coord_polar("y", start=0) ```### Small multiplesWe might want to compare frequencies across multiple variables, for example```{r}## lets create a frequency table of the number of breeds mentioned in studes for each herbivore groupgrouped <- data |>select(contains("vine")) |>pivot_longer(names_to ="herbivore", values_to ="breed", cols =1:6) |>group_by(herbivore) |>count(breed)grouped``````{r}#| label: stacked barsgrouped |>drop_na() |>ggplot(aes(herbivore, n, fill = breed)) +geom_col(position ="fill") +scale_y_continuous(labels = scales::percent_format())``````{r}#| label: faceted (small multiples)grouped |>drop_na() |>ggplot(aes(reorder(breed, n), n)) +geom_col(position ="dodge") +coord_flip() +labs(x ="", y ="No of studies") +#scale_y_continuous(labels = scales::percent_format()) +facet_wrap(~herbivore) ```
