Level Up 02: Manipulating Factors

# Level Up 02: Manipulating Factors

---

# SeminRs

- Informal, optional weekly sessions to help build a 'portfolio of skills'

- 1-2 hours of instruction, demos, walk throughs & activities to try out

### *Essentials*

- Focused on the fundRmental skills to help you get started with learning R  
- Covering: basic wrangling & visualising data, '.bold.pink[pretty]' R Markdown, inline code, debugging...

### *Level Up*

- Focused on more advanced programming skills & applying these skills to new .bold.orange[fun] topics  
- Covering: Papaja, advanced wrangling & manipulation of data, '.bold.italic.pink[even prettier]' R Markdown, spotifyR...

<br>

*Session topics are not fixed - use the Padlet linked on Canvas for suggestions!*

---

# Setup & Suggested Workflow

- Create one R project file for all seminR sessions

- Within this directory, create an r_docs & data folder & save all Rmds and datasets to these folders respectively

- Make a cheat sheet of useful functions and .orange[#] comment their meaning and usage as you go through seminRs, practicals, tutorials etc.

#### .orange[Reminder]:

File > New Project... New Directory > New Project > *Give your project a name & location*

File > New File > R Markdown...  *Remember to save this file in your r_docs folder!*

### .orange[Task]: 
Open your seminRs project, create a new Rmd file & load tidyverse & palmerpenguins

---

# Session Objectives

### Wrangling Factors with forcats

- Understand factors & categorical variables

- Learn & practise some common functions (from base R, dplyr, forcats) to manipulate factors

To check your answers or follow along, [download the solutions Rmd](https://and.netlify.app/seminr/02/level_up/levelup_02.Rmd)

---

# Factors & Categorical Data

- Factors represent categorical data  
- Numeric values are used to represent different levels  
- They have a fixed, known set of possible values  
- forcats is a useful package when working with factors, loads with library(tidyverse) or library(forcats)

### Examples

- Marital status 
- Occupation
- Education level
- Eye colour 
- Pretty much anything...

---

# Working with Factors in R

### .orange[Task 1]:

Read in the dataset below

```r
spotify_data <- readr::read_csv("../data/spotify_decades_data.csv") # if saved already
spotify_data <- readr::read_csv("https://raw.githubusercontent.com/de84sussex/DS_spotify/main/spotify_decades_data.csv")
```

### .orange[Task 2]:

Explore the dataset, is there anything odd or unusual?

---

# Creating/Converting Factors

IMO, easiest way to convert from an existing variable is to use a combination of factor() from base R and mutate() from dplyr

#### Step 1: add a pipe following reading the data into mutate()

#### Step 2: specify the *new* column name

#### Step 3: use factor() on the column you want to change, specify the labels for each category

```r
spotify_data <- readr::read_csv("../data/spotify_decades_data.csv") %>% 
  mutate(decade_fct = 
           factor(decade_fct, 
                  labels = c("70s", "80s", "90s", "00s")))
```

### .orange[Task]:

Follow these steps and update your code to read in decade_fct as a factor

---

# Inspecting Factors

### Base R

class() for seeing/checking the data type is correct

levels() to see the levels of a factor

table() for table of groups & counts

### forcats

fct_count() gives factor level and counts, similar to table()

### .orange[Task]:

Try these functions on our decade_fct variable, remember we subset a variable by using the .orange[$] 
dataset$column

---

# Filter For Demo

```r
# adding in filter to see differences between groups more clearly in following tasks

spotify_data <- spotify_data %>% filter(.data = ., track.popularity > 50 & track.explicit == FALSE)
```

---

# Changing Factor Ordering

fct_relevel() for reordering the factor levels, useful if factors are ordered (i.e. low, med, high) and they're originally set up inappropriately (i.e. low = 1, high = 2, med = 3)

```r
dataset$column <- fct_relevel(dataset$column, c("low", "medium", "high")) 
```

fct_infreq() reorders the levels so the most frequent category becomes level 1, second most frequent becomes 2, and so on, useful with plots

```r
plot <- ggplot(data, aes(x = fct_infreq(categorical_variable))) + 
  geom_bar() + 
  coord_flip()
```

### .orange[Task]:

Relevel the decade_fct variable so that 00s is level 1, 90s is level 2, 80s is level 3, and 70s is level 4  
Create a plot of decade_fct which is reordered so the bars are displayed in order of their frequency

---

# Changing Factor Levels

fct_recode() changes the level names/values

```r
dataset$column <- fct_recode(dataset$column, new_name_1 = "old_name_1", new_name_2 = "old_name_2", new_name_3 = "old_name_3", new_name_4 = "old_name_4")
```

fct_collapse() collapses multiple levels into one

```r
dataset$column <- fct_collapse(dataset$column, new_name_1 = c("old_name_1", "old_name_2"), new_name_2 = c("old_name_3", "old_name_4"))
```

### .orange[Task]:

Recode the levels so that 00s is naughties, 90s is nineties, 80s is eighties, 70s is seventies  
Collapse the levels so that seventies & eighties are called oldies, & nineties & naughties are called newbies

---

# For extRa fun

Try using some of the commands we've used today to wrangle another dataset!

```r
peng_data <- palmerpenguins::penguins # to load the data
```

### .orange[Task]:

1. Inspect the species factor using class(), levels(), & fct_count()  
1. fct_reorder() the levels of species so that the species with the highest count is level 1, and the species with lowest count is level 3 (you'll need to look at the output from fct_count to find out the *n*s & which order to put them into fct_relevel())  
1. Google images of the 3 species, decide which one is cute, which one is cuter, and which one is the cutest
1. fct_recode() the levels to reflect your cuteness ratings given to the different species  
1. fct_collapse the cute and cuter levels into one category called just_cute

---