Loading the Libraries and Data

library(tidyverse) # who doesn't want to be tidy
library(gt) # nice tables
library(ggthemes) # more themes for ggplots
library(RColorBrewer) # color scales
library(here) # consistent method for working in file structure

Loading the TidyTuesday in the normal way.

tuesdata <- tidytuesdayR::tt_load(2023, week = 30)

scurvy <- tuesdata$scurvy

Reformat and Clean the data

Take a look at the data.

scurvy %>% gt()

Small data set, maybe best as a table.

First some cleanup. First, I’m replacing all the underscores (“_”) in the treatment descriptions with spaces. I use str_replace_all() from stringr inside a mutate() for that.

scurvy <- scurvy %>%
  mutate(treatment = str_replace_all(treatment, "_", " "))

Reshaping the Data with pivot_longer()

A heatmap is like a table, but a bit more visually interesting. I’d like to plot the treatment vs. the symptoms. To do so, the data needs to be reshaped into the long format. This is easily done using pivot_longer() from tidyr. Here, I give it the name of the two new columns, symptom and severity. The symptom column will contain the values gum rot, skin sores, etc. that are currently fields. The severity column will contain the likert scale rating for that symptom (1- mild, 2- moderate, 3- severe). I’m also going to split those ratings into a numerical (1, 2, 3) and descriptive (“mild”, “moderate”, “severe”) rating using separate() from tidyr.

scurvy2 <- scurvy %>%
  pivot_longer(4:7, names_to = "symptom", values_to = "severity") %>%
  separate(
    severity,
    sep = "_",
    into = c("Numerical_Severity", "Descriptive_Severity")
  )

Now, I want to get rid of the day 6 label (d6) from the symptom. There are no other time points, so the data is somewhat extraneous. Here I’m using separate_wider_delim(). This is actually the active/ recommended separate function in tidyr. Both separate() and extract() still work, but it is recommended that users switch to the new family of separate functions. Instead of into for the new column names, this function uses names. The list of names can contain NA for sections that you wish to discard. Here, I throw away the second section (_d6). Then I use str_replace_all() to clean out the rest of the underscores in the symptom names. Note that if I did this step first, I’d need a different delimiter. I generally like to do any separations before I do str_replace() because you are more likely to have a unique separator/delimiter before you clean up special symbols.

scurvy2 <- scurvy2 %>%
  separate_wider_delim(symptom, delim = "_d6", names = c("symptom", NA)) %>%
  mutate(symptom = str_replace_all(symptom, "_", " "))

Making the Heatmap

I never remember the color palette names, so here they are. The par(cex = 0.5) command is a base R method of changing the parameters for the next plot. The cex parameter controls the text size. There are a lot of palettes, so if you don’t do this the labels will be squished.

par(cex = 0.5) 
display.brewer.all()  

ggplot(scurvy2, aes(treatment, symptom)) +   
  geom_tile(aes(fill = Numerical_Severity))  +
  geom_text(aes(label = Descriptive_Severity)) +
  scale_fill_brewer(palette = "YlOrRd")

scurvy_heatmap <- ggplot(scurvy2, aes(symptom, study_id)) +   
  geom_tile(aes(fill = Numerical_Severity))  +
  geom_text(aes(label = Descriptive_Severity)) +
  scale_fill_brewer(palette = "YlOrRd") +
  theme(legend.position = "none") 

scurvy_heatmap

scurvy_heatmap <- scurvy_heatmap +
  scale_y_continuous(breaks = 1:12, labels = scurvy$treatment)

scurvy_heatmap

scurvy_heatmap <- scurvy_heatmap +
  theme_void() +
  theme(axis.text = element_text(size = 10),
        axis.title = element_text(face = 'bold')) +
  labs(title = "James Lind's Study of Scurvy Treatments- 1757", 
       caption = "data from medicaldata R package",
       subtitle = "Results at Day 6") +
  ylab('Treatments') + xlab('Symptoms') +
  theme(axis.text.x = element_text(angle = 45))
  
scurvy_heatmap

ggplot(scurvy2, aes(treatment, symptom)) + 
  geom_tile(aes(fill = Numerical_Severity)) +
  theme_void()

scurvy_heatmap <- scurvy_heatmap +
theme(legend.position = "none") +
  theme(axis.title.y = element_text(angle = 90))

scurvy_heatmap

ggsave(here("post", "2023-07-tidytuesday-scurvy", "thumbnail.png"))

Saving 7 x 5 in image

Thoughts about Scurvy

Despite having experimental evidence that fresh citrus was effective in treating scurvy, James Lind recommended other treatments and preparations in his book. Of particular note is a concentrate of citrus juices that could be bottled and stored indefinitely. Vitamin C is very unstable; the processing and storage removed almost all the vitamin C from the juice. This is covered in a fascinating paper by R. E. Hughes, where he prepared and measured the vitamin C levels in various treatments recommended by Lind. Hughes does touch on the issue of cider. Lind’s data seems to suggest that cider was moderately effective as a treatment. Fermentation and aging does degrade vitamin C, so the cider likely did not have more than trace amounts of vitamin C. I suspect there was still some and perhaps enough to mitigate some of the worst impacts of scurvy.