Data Reporting with R: Best Practices from Tables to Time Series

Author

Kaburungo

Demographic Pyramids

Load packages

This requires R packages for data manipulation, data visualization and relative file paths

# Load packages
if(!require(pacman)) install.packages("pacman")

Loading required package: pacman

pacman::p_load(tidyverse, # to clean, wrangle, and plot data
       here,      # to locate files
       apyramid)   # package dedicated to creating age pyramids

Data Preparation

Intro to the Dataset

We’ll be using a simulated HIV dataset of linelist cases in Zimbabwe during 2016.
Our focus : age relate and sex variables for the demographic pyramid.

Importing data

Let’s Import our dataset into Rstudio and inspect the variables:

# Import the data from CSV file
hiv_data <- read.csv(here("C:/Users/Perminus Njiru/OneDrive - LVCT Health/Desktop/Freecodecamp/R/Care_and_Treatment_Analysis/Input/hiv_zw_linelist_2016.csv"))

# display the data frame
head(hiv_data)

  age_group    sex hiv_status
1     20-24 female   positive
2     35-39   male   positive
3     15-19 female   negative
4     40-44   male   negative
5     45-49   male   positive
6     35-39 female   negative

Our dataset has 28000 cases across 3 columns : age_group, sex, and hiv_status.
To create a demographic pyramid of HIV Prevalence, we’ll filter the data to include only HIV Positive individuals.

#filter and save Hiv Pos as a new dataset
hiv_case <- 
  hiv_data %>% 
  filter(hiv_status == "positive")
# view dataset
head(hiv_case)

  age_group    sex hiv_status
1     20-24 female   positive
2     35-39   male   positive
3     45-49   male   positive
4     50-54   male   positive
5     20-24 female   positive
6     45-49 female   positive

Data Inspection

Next, we’ll look at a summary table for age_group and sex to check that our data is clean.

# summarize the data into a table by age group and sex
hiv_case %>% 
  count(age_group,sex) %>% head(5)

  age_group    sex  n
1     00-04 female  3
2     00-04   male  6
3     05-09 female  8
4     05-09   male 10
5     10-14 female 42

Age groups are in ascending order - this is crucial for plotting the demographic pyramid

Grouping and aggregating Data

To plot the pyramid bars, we need to summarize the linelist observation into an aggregated table.
We want positive females totals and negative male totals, to plot bars on opposite sides of the axis.
We can use count() and mutate() to get total cases and percents grouped by age group and sex.

# Create new subsets with grouped counts
pyramid_data <- 
  hiv_case %>% 
  #count total cases by age group and gender
  count(age_group,sex,name = "total") %>%
  # create new columns for x axis values on the plot
  mutate(
    #add column with diverging axis values - converts male counts to negative 
    axis_count = ifelse(sex == "male",-total,total),
    #add column with percentage axis values
    axis_percent = round(100 * (axis_count/sum(total)),digits = 1)
    )
head(pyramid_data)

  age_group    sex total axis_count axis_percent
1     00-04 female     3          3          0.1
2     00-04   male     6         -6         -0.2
3     05-09 female     8          8          0.2
4     05-09   male    10        -10         -0.3
5     10-14 female    42         42          1.1
6     10-14   male    23        -23         -0.6

Now that the data is summarized in appropriate format, we can use pyramid_data to create population pyramid with {ggplot2}!

Alternative to {ggplot2=“” ()=““}

The {apyramid} package can be a useful tool that contains a function age_pyramid(), which allows for the rapid creation of age-sex pyramids:

# Start with the original linelist (no need to summarize counts)
hiv_case %>% 
  mutate(age_group = factor(age_group)) %>% 
  apyramid::age_pyramid(
    # Required arguments:
    age_group = "age_group",
    split_by = "sex")

Plotting Demographic Pyramids with {ggplot2()}

Remember; a demographic pyramid is a modified version of a stacked bar plot
A basic stacked bar plot with geom_col () needs a categorical variable against a continuous variable (e.g, age_group vs. total), and fill set to a second categorical variable (e.g. sex)

# Basic stacked bar plot: bars stacked on top of each other
# initialize plot
ggplot() +
# Create bar graph using geom_col()
  geom_col(data = pyramid_data, # specify dataset for plotting
           aes(x = age_group,   # Indicate categorical x variable
              y = total,        # Indicate continuous y variable
              fill = sex)) +    # fill by second categorical variable
# Modify theme
  theme_light()+
  coord_flip()

Using geom_col() for demographic pyramids

We can build on the stacked bar code above to create a demographic pyramid
This time we use axis_counts for the y variables instead of the total, which has negative male counts

# Create and save plot to environment
demo_pyramid <- 
  ggplot() +
  geom_col(data = pyramid_data,
           aes(x = age_group,
               y = axis_count, # indicate special y variable
               fill = sex)) +
  theme_light() +
  # Flip x and y axes
  coord_flip()
demo_pyramid

We can also use percentage values ( axis_percentage) on the y axis

demo_pyramid_percentage <- 
  ggplot() +
  geom_col(data = pyramid_data,
           aes(x = age_group,
               y = axis_percent, # indicate special y variable
               fill = sex)) +
  theme_light() +
  # Flip x and y axes
  coord_flip()
demo_pyramid_percentage

Plot Customization

Next we’ll re-scale the case count axis, add informative labels, and edit non-data elements
The current graph shows asymmetrical axis limits due to varying case count between genders

Choosing the limits

Our goal is an axis that is symmetrical, with the same length on both sides around zero.

# store highest case as an object
max_count <- 
  max(pyramid_data$total)
max_count

[1] 318

We will edit our continuous axis using the scale_y_continuous () function
Set the upper and lower axis limits to max_count.
Define the spacing between axis breaks.
Convert the negative labels left of 0 to their absolute values (remove minus sign)

# Add scales layers to the previous graph
custom_axes <- 
  demo_pyramid +
  # Scale function for y axis (axis_count)
  scale_y_continuous(
    # Specify upper and lower limits of the same length, for symmetry
    limits = c(-max_count,max_count),
    #Specify the spacing between axis break labels
    breaks = scales::breaks_width(100),
    # Make axis numbers absolute so male counts appear positive
    labels = abs
  )
custom_axes

Equal upper and lower limits on both sides ensures a symmetrical and accurate visual comparison.
Negative numbers from axis_count are converted to positive numbers using labels = abs
The width of breaks are fixed at a specific value using breaks_width(100)

Watch Out

Although it looks like the X axis, axis_counts is actually the y axis
The code to create demo_pyramid had age_group mapped to x and axis_count mapped to y
we then flipped the orientation of the plot using coord_flip().

demo_pyramid <- 
  ggplot() +
  geom_col(data = pyramid_data,
           aes(x = age_group,
               y = axis_count, # indicate special y variable
               fill = sex)) +
  theme_light() +
  # Flip x and y axes
  coord_flip()

Add Custom Labels

Use the labs() functions to informative main title, subtitle, axis titles and caption.

custome_labels <- 
  # Starts with previous demographics pyramid
  custom_axes +
  #labs() functions controls label text
  labs(
    title = "HIV Positive cases by Age and Sex",
    subtitle = "Zimbabwe (2016)",
    x = "Age Group",
    y = "Incidences",
    fill = "Sex",
    caption = stringr::str_glue("Data are from simulated linelist \nn = {nrow(hiv_case)}")
  )
custome_labels

Enhance Color Scheme and Themes

Let’ apply some style adjustment to our demographic pyramid

Custom_color_theme <- 
  # Build upon the previous plot
  custome_labels +
  # Manually specify custom colors for each sex
  scale_fill_manual(values = c("female" = "lightblue",
                               "male" = "darkblue"),
                    #capitalize legend labels
                    labels = c("Female","Male")) +
  #fine - tune theme elements for a cleaner look
  theme(axis.line = element_line(colour = "black"), # make axis lines black
         # Center title and subtitle
         plot.title = element_text(hjust = 0.5),
         plot.subtitle = element_text(hjust = 0.5),
         # Format caption text
         plot.caption = element_text(hjust = 0, #align left
                                     size = 11, #increase font size
                                     face = "italic")) #italicize text
Custom_color_theme

Wrap-Up

Demographic pyramids are vital tools for visualizing disease distribution by age and sex.
The techniques used here can be extended to other graphs with negative and positive values.
You can now take these concepts to visualize cases against baseline populations or the impact of health interventions.
This knowledge is invaluable for epidemiological analysis and reporting.