Factors

1 Presentation

You can download the course slides for this section here

Getting Started

• Make sure that you are working in your course project
• Create a new quarto document and name it “forcats.qmd”
• Insert a code chunk and load 2 important libraries
• Insert a new code chunk- Write source(here("scripts", "01_import.R")) in the chunk
• Discuss what the code does and run it
• Write a short headline to each code chunk
• Change the YAML header to style your document output.

The YAML header can look like this

---
title: "TITLE"
subtitle: "SUBTITLE"
author: "ME"
date: today
format: 
  html:
    toc: true
    toc-depth: 2
    embed-resources: true
    number-sections: true
    number-depth: 2
    code-fold: true
    code-summary: "Show the code"
    code-tools: true
execute:
  message: false
  warning: false
---

# Load libraries
library(tidyverse)  # Data wrangling and plots
library(here)       # File control in project

# Load the cleaned soldiers dataset
source(here("scripts", "01_import.R"))

2 forcats

 

Use the soldiers dataset for the following exercises.

race

race is now ordered alphabetically (try: soldiers |> count(race)))

• fix race so that it is ordered after frequency (fct_infreq())
• Use a barplot to show the number of soldiers within each race (aes(x = race)) ::: {.cell}

soldiers |> 
  mutate(
    race = fct_infreq(race)
  ) |> 
  
  ggplot(aes(x = race, fill = race))+
  geom_bar(show.legend = FALSE)+
  scale_x_discrete(labels = scales::label_wrap(8)) # one of many ways to fix long labels on the axis

:::

race

• Update your 01_import.R file so that race will be ordered after frequency when you source the script
• Remember to comment your code

Ethnicity

Background information: DODRace was collected by assigning fixed race values (1:7) to each soldier. Ethnicity was a free text field where the soldiers themselves have filled out their race.

• Explore Ethnicity using count() and view() ::: {.cell}

soldiers |> 
  count(Ethnicity) # add  |> view()

:::

 

fct_lump()

OMG.. we probably need to merge some of the many Ethnicity groups.
Try with fct_lump()

• Can you find a number of levels that’s reasonable? ::: {.cell}

soldiers |> 
  mutate(
    Ethnicity = fct_lump(Ethnicity, 5) # I dont think this is reasonable see below
  ) |> 
  count(Ethnicity)

:::

fct_collapse()

Hmmm… fct_lump() is probably not the best choice for the Ethnicity variable. It has too many groups, and many groups have similar sounding names. We need to fix this manually.

• Here is a code example where we merge all the Apache.
• Merge a few more groups together by building on this code.
• Add the code to your 01_import.R file

soldiers |> 
  mutate(
    Ethnicity = fct_collapse(Ethnicity,
      Apache = c("Apache Blackfoot",
                 "Apache Blackfoot Cherokee Crow",
                 "Apache Cherokee",
                 "Apache Kiowa Mexican",
                 "Apache Mexican" )
    )
  ) 

# Ethnicity is really long, and it might be better to fix it with a case_when() 
# and functions from the stringr package
# The function str_detect() looks for pattern in a text file and returns TRUE
# if the pattern is present

soldiers |> 
  mutate(
    Ethnicity2 = case_when(
      str_detect(Ethnicity, "Apache") | str_detect(Ethnicity, "Cherokee")  ~ "Native American", 
      str_detect(Ethnicity, "Cuban") | str_detect(Ethnicity, "Dominican") | str_detect(Ethnicity, "Caribbean") ~"Caribbean", 
      is.na(Ethnicity) ~ NA, 
      .default = Ethnicity
    ), 
    .keep = "used"  # Only keep the new variables and the ones we used to create them. For ease of comparison
    ) |> 
  drop_na() # Remove all the rows with an NA value. For ease of comparison

category

DODRace, race, and Ethnicity are all true factors in the sense that no values in any of these variables are more ´race´ than other values. Think about category do the values here imply more or less of the same thing?

• Fix category by changing it into an ordered variable. Use the function factor() and set the argument ordered = TRUE
• Add the code to your 01_import.R file
• Try to recreate some of the plots we did on day 2 were we colored the points using category - Notice the difference? ::: {.cell}

soldiers |> 
  mutate(
    category = factor(category,
                      levels = (c("Underweight",
                                "Normal range",
                                "Overweight",
                                "Obese")),
                      ordered = TRUE)
  ) 

:::

 

Improve your home assignment

• Go back to your home assignment and see if you can improve anything using your new factor and forcats skill
• Remember to change project (top right corner in Rstudio)
• Using the menu in the top right corner, you can switch between your course project and your home assignment

 

2.1 Challenges with forcats + ggplot + dplyr

In mpg

• Collapse the model types a4 quattro and a4 into a4
• Create a bar plot sorted by frequency
• Work out a way to show what manufacturers the models belong to

Here is my suggestion

I have cheated and used some functions (str_to_title() and facet_grid()), and theme options that I haven´t showed you before. ::: {.cell}

mpg |> 
  group_by(manufacturer) |> 
  mutate(
    model = model |> str_to_title() |>  fct_collapse(A4 = c("A4", "A4 Quattro")) |> fct_infreq(),
    manufacturer = str_to_title(manufacturer)) |> 
  ggplot(aes(y = model, fill = manufacturer)) +
  geom_bar()+
  geom_text(aes(x = 0.5, label = model), size =3,  hjust = 0, check_overlap = TRUE)+ # Display the model name at the position (x=0.5, y = model)
  scale_x_continuous(expand = c(0,0))+ # Remove the padding between the y-axis and the start of the bars
  facet_grid(rows = vars(manufacturer), 
             scales = "free_y", 
             space = "free_y",
             switch = "y")+
  ggthemes::theme_pander()+
  theme(axis.text.y=element_blank(), # Remove the names from y-axis (we used geom_tect instead)
        axis.ticks.y = element_blank(), # Remove y axis ticks (the small lines)
        strip.text.y.left = element_text(angle = 0, hjust = 1), # Change strip text orientation
        legend.position = "none" # remove fill legend
        ) +
  labs(
    title = "Count of car models in `ggplot2::mpg` data set",
    x = "Count of car models",
    y = NULL,
    caption = "Consider a different fill color scale. The current one seems to imply a gradient"
  )

:::

Fix this starwars plot

• Order the columns after frequency
• Pimp it with a nice theme and consider adding some nice colors
• Correct the y-axis label and give the plot a title
• Consider lumping together some of the levels ::: {.cell}

ggplot(starwars, aes(y = eye_color)) + 
  geom_bar() 

:::

Here is my suggestion

I have cheated and used three functions (str_to_title(), after_stat(), and scale_fill_gradient()), that I haven´t showed you before.
::: {.cell}

starwars |>
  mutate(eye_color = fct_recode(str_to_title(eye_color))) |>  # Change all factor levels to Title case
  mutate(eye_color = fct_lump(eye_color, 7) |> fct_infreq()) |> 
  ggplot(aes(y = eye_color,
             fill = after_stat(count))) +                      # Set the fill color to the count value
  geom_bar() +
  ggthemes::theme_foundation()+
  scale_fill_gradient(low = "grey", high = "black")+           # Create a new fill scale going from grey to black
  labs(
    x = "Count", 
    y = "Eye Color",
    title = "Eye color counts of Starwars characters",
    caption = "Consider... Is the grey gradient disturbing? e.g. ´brown´ has a black color ")+
  theme(plot.title.position = "plot")   # Place the title all the way to the left side

:::

In table1 (another inbuilt dataset)

table1 displays the number of TB cases documented by WHO in Afghanistan, Brazil, and China between 1999 and 2000.

• Create a new variable called case_pr_mill_pop (cases pr. million)
  
• recode country labels so that (Afghanistan = Afg, Brazil = Bra, China = Chi)
  
• reorder country factor after case_pr_mill_pop
  
• arrange the table after country

table1 |>
  mutate(
    case_pr_mill_pop = cases/(population/1e6),
    country = fct_recode(country,
                         "Afg" = "Afghanistan",
                         "Bra" = "Brazil",
                         "Chi" = "China"),
    country = fct_reorder(country, case_pr_mill_pop)) |> 
  arrange(country)