EDA w/NMU enrollment data

gganimate

Create animated .gifs from data.

# load and clean_names NMU data
nmu <- read.csv("http://euclid.nmu.edu/~joshthom/Teaching/DAT309/Week6/10%20BOS%2010%20Year%20History%20Deidentified%2010.3.23.csv", sep  = "\t")
nmu <- janitor::clean_names(nmu)

# install.packages("gganimate")
# library(gganimate)

  ggplot(nmu, aes(x = department, fill = gender)) + 
    geom_bar() +
      theme(axis.text.x = element_text(angle = 60, hjust = 1 )) +
    labs(title = 'Year: {closest_state}', 
                  x = 'Departments', 
                  y = 'Enrollment') + 
    transition_states(states=year,
                      transition_length = 1,
                      state_length = 4) +
    enter_fade() + 
    exit_fade() + 
    ease_aes('linear')
  
  last_animation()
  anim_save("anim_save("~/Google Drive/Teaching/DAT309/Week7/nmu_10yr_all.gif)

View the animation [here] ## Facets (10.4)

Two ways to make subplots, facet_wrap() and facet_grid(). For facet_grid, the first argument is again a function but in the syntax var1_row ~ var2_col.

suppressPackageStartupMessages(
  library(tidyverse)
)
ggplot(nmu, aes(x = year)) + 
  geom_bar() + 
  facet_grid(housing ~ class_level) +
  scale_x_continuous(breaks = seq(2013, 2023, 1)) +
  theme(axis.text.x = element_text(angle = 60, hjust = 1)) +
  labs(title = "NMU Enrollment - On campus & Off")

I don’t like the ordering of class level, let’s fix it.

nmu <- mutate(nmu,class_level = 
                factor(class_level, 
                       levels = c("Freshman","Sophomore",
                                  "Junior","Senior",
                                  "Graduate","Vocational")))

ggplot(nmu, aes(x = year)) + 
  geom_bar() + 
  facet_grid(housing ~ class_level) +
  scale_x_continuous(breaks = seq(2013, 2023, 1)) +
  theme(axis.text.x = element_text(angle = 60, hjust = 1)) +
  labs(title = "NMU Enrollment - On campus & Off")

Change Y to Yes, and blank to No

nmu <- mutate(nmu,first_generation = if_else(first_generation == "Y","Yes","No"))

To zoom in we “free up” the scales from their locked default position.

ggplot(nmu, aes(x = year)) + 
  geom_bar() + 
  facet_grid(first_generation ~ class_level) +
  # explore scales = "free" above
  scale_x_continuous(breaks = seq(2013, 2023, 1)) +
  theme(axis.text.x = element_text(angle = 60, hjust = 1)) +
  labs(title = "NMU Enrollment - First Generation On Campus & Off")

Exercise

1. Experiment with the parameter nrow in facet_grid.

2. Experiment by changing the formula to . ~ class_level and class_level ~ .

Exploratory Data Analysis w/ NMU data

Identify enrollment by housing (On campus or Off)

# count enrollment by housing
nmu_housing_year <- nmu |> 
  group_by(year, housing) |> 
  summarize(n = n()) |> 
  ungroup()

# pivot so that rows correspond to years
year_vs_housing <- nmu_housing_year |> 
  pivot_wider(
    names_from = housing, 
    values_from = n) |> 
  janitor::clean_names()

Compute year-to-year differences for each class

year_vs_housing <- mutate(year_vs_housing, 
        on_campus_diff = c(NA,diff(year_vs_housing$on_campus)))
year_vs_housing <- mutate(year_vs_housing, 
        off_campus_diff = c(NA,diff(year_vs_housing$off_campus)))

Pivot the differences into one variable & pivot their names into another variable.

nmu_diffs <- year_vs_housing |> 
  pivot_longer(
    cols = contains("diff"), 
    names_to = "housing_diffs", 
    values_to = "diffs")

Plot year to year differences for each housing.

ggplot(nmu_diffs, 
       aes(x = year, y = diffs, 
           group = housing_diffs, 
           color = housing_diffs, 
           fill = housing_diffs, 
           )) + 
  geom_bar(position = "dodge", stat = "identity") +   
  scale_x_continuous(breaks = seq(2013, 2023, 1)) +
  scale_fill_discrete(labels = c("Off Campus","On Campus")) +
  scale_color_discrete(labels = c("Off Campus","On Campus")) +
  theme(axis.text.x = element_text(angle = 60, hjust = 1)) +
  labs(fill = "Differences", color = "Differences") +
  geom_hline(yintercept = 0)

Exercises

1. Vary the geom usage above. Which is the most useful?

2. Repeat the anaylsis above for gender instead of class_level

3. Add a position argument (eg, "identity", "fill", "dodge) to the geom above.