2020 Durham City and County Resident Survey

Application exercise

The main questions we’ll explore today are:

“What are the demographics and priorities of City of Durham residents?”
“How do City of Durham residents feel about quality of services by the city and how, if at all, does that vary by income?”

Goals

Wrapping up discussion on categorical data
Visualizing and summarizing numerical data
Improving visualizations for visual appeal and better communication

Packages

library(tidyverse)

Data

The data for this case study come from the 2020 Durham City and County Resident Survey.

First, let’s load the data:

durham <- read_csv("data/durham-2020.csv")

Rows: 803 Columns: 49
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr  (7): please_define_other_32_5, primary_language, please_define_other_34...
dbl (42): id, overall_quality_of_services_3_01, overall_quality_of_services_...

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

In the last application exercise we manipulated the data a bit.

durham <- durham |>
  rename(
    own_rent = do_you_own_or_rent_your_current_resi_31,
    income = would_you_say_your_total_annual_hous_35,
    # a new one
    qos_city = overall_quality_of_services_3_01
  ) |>
  mutate(
    income = as_factor(income),
    own_rent = as_factor(own_rent)
  )

Visualizing relationships between categorical variables

Exercise 1

Visualize and describe the relationship between income and home ownership of Durham residents.

Stretch goal: Customize the colors using named colors from http://www.stat.columbia.edu/~tzheng/files/Rcolor.pdf.

Add your answer here.

# add code here

Exercise 2

Calculate the proportions of home owners for each category of Durham residents. Describe the relationship between these two variables, this time with the actual values from the conditional distribution of home ownership based on income level.

Add your answer here.

# add code here

Exercise 3

Recode the levels of these two variables to be more informatively labeled and calculate the proportions from the previous exercise again.

# add code here

Visualizing numerical data

Exercise 4

One of the questions on the survey is “How satisfied are you with the overall quality of services provided by the city?” A response of 1 indicates Very Dissatisfied and a response of 5 indicates Very Satisfied. The responses for this question are in the variable qos_city. This could be considered an ordinal, categorical variable but can also be treated as a numerical variable in an analysis. Let’s do the latter!

Visualize and describe the distribution of qos_city. If you get a warning with your visualization, comment on what it means.

Add your response here.

# add code here

Exercise 5

Calculate the mean and median of the distribution of qos_city. If these values are not exactly the same, can you explain what the difference might be attributed to?

Add your response here.

# add code here

Exercise 6

Based on the shape of the distribution of qos_city, which measure of spread (variability) is more appropriate? Calculate that value and interpret it in context of the data.

# add code here

Exercise 7

Make a box plot of qos_city and comment on how the values you calculated map on to the box plot.

# add code here

Exercise 8

How the average level of happiness with quality of services provided by the city vary by income group?

Stretch goal: Visualize mean qos_city by income group.

Add your response here.

# add code here

Recap

Conceptual

Some of the terms we introduced are:

Marginal distribution: Distribution of a single variable.
Conditional distribution: Distribution of a variable conditioned on the values (or levels, in the context of categorical data) of another.

R

In this application exercise we:

Defined factors – the data type that R uses for categorical variables, i.e., variables that can take on values from a finite set of levels.
Reviewed data imports, visualization, and wrangling functions encountered before:
- Import: read_csv(): Read data from a CSV (comma separated values) file
- Visualization:
  - ggplot(): Create a plot using the ggplot2 package
  - aes(): Map variables from the data to aesthetic elements of the plot, generally passed as an argument to ggplot() or to geom_*() functions (define only x or y aesthetic)
  - geom_bar(): Represent data with bars, after calculating heights of bars under the hood
  - geom_histogram(): Represent data with a histogram
  - geom_boxplot(): Represent data with a box plot
  - geom_point(): Represent data with points
  - labs(): Label x axis, y axis, legend for color of plot, title` of plot, etc.
- Wrangling:
  - mutate(): Mutate the data frame by creating a new column or overwriting one of the existing columns
  - count(): Count the number of observations for each level of a categorical variable (factor) or each distinct value of any other type of variable
  - group_by(): Perform each subsequent action once per each group of the variable, where groups can be defined based on the levels of one or more variables
  - summarize(): Calculate summary statistics
Introduced new data wrangling functions:
- rename(): Rename columns in a data frame
- as_factor(): Convert a variable to a factor
- drop_na(): Drop rows that have NA in one ore more specified variables
- if_else(): Write logic for what happens if a condition is true and what happens if it’s not
- case_when(): Write a generalized if_else() logic for more than one condition
- fct_relevel: Change the order of levels in a factor
Introduced new data visualization functions:
- geom_col(): Represent data with bars (columns), for heights that have already been calculated (must define x and y aesthetics)
- scale_fill_viridis_d(): Customize the discrete fill scale, using a color-blind friendly, ordinal discrete color scale
- scale_y_discrete(): Customize the discrete y scale
- scale_fill_manual(): Customize the fill scale by manually adjusting values for colors

Quarto

We also introduced chunk options for managing figure sizes:

fig-width: Width of figure
fig-asp: Aspect ratio of figure (height / width)
fig-height: Height of figure – but I recommend using fig-width and fig-asp, instead of fig-width and fig-height

Acknowledgements

This dataset was cleaned and prepared for analysis by Duke StatSci PhD student Sam Rosen.