11b-plots-examples-tornado.Rmd

### Tornado Plot
#### Overview
This article will demonstrate how to construct a tornado plot in `ggplot2`. A tornado plot is useful for comparing counts (or percents) of events in a back-to-back fashion. They can highlight equalities or imbalances by symmetry between treatment arms and also provide a view of each distribution separately.

#### Packages and Sample Data
Let's start with accessing ADAE from the PhUSE Test Data Factory repository.

```{r, message = FALSE, warning = FALSE, error = FALSE}
# Packages
library(ggplot2)
library(dplyr)
library(ggh4x)

# Data
adae <- haven::read_xpt(
  paste0("https://github.com/phuse-org/TestDataFactory/",
         "raw/main/Updated/TDF_ADaM/adae.xpt"))
```

#### Data Wrangling
In order to make this example meaningful, data wrangling and processing steps will be covered here. In practice, you might have other considerations or processing steps that need to be considered.

The first step is to prepare the data for wrangling. For this example, we will limit our focus on only 2 of 3 treatment arms and select relevant variables for tabulation.
```{r message = FALSE, warning = FALSE, error = FALSE}
start <- adae %>%
  select(TRTA, USUBJID, AESEV, AEBODSYS) %>%
  filter(TRTA != "Placebo")
```

The next step is to tabulate unique counts, per subject, of the adverse event (AEBODSYS) by severity combinations.
```{r message = FALSE, warning = FALSE, error = FALSE}
event_counts <- start %>%
  distinct(USUBJID, AEBODSYS, AESEV, .keep_all = TRUE) %>%
  count(TRTA, AEBODSYS, AESEV)
```

#### Basic Plot
Now that we have some data, we can start with a basic tornado plot. We will rely on the `facetted_pos_scales()` function from the `{ggh4x}` package to help position scales.
```{r message = FALSE, warning = FALSE, error = FALSE, fig.width=12, fig.height= 8}
ggplot(event_counts, aes(fill = AESEV, y = n, x = reorder(AEBODSYS, n))) +
  geom_bar(position="stack", stat="identity") +
  coord_flip() +
  facet_wrap(~ TRTA, scales = "free_x") +
  facetted_pos_scales(y = list(
    scale_y_reverse(),
    scale_y_continuous())
  )
```

#### Customized Plot
We have a few customization to be included in order to make this plot more presentable. Specifically, we will:

1. Specify a custom fill palette
2. Set the y-scales of each individual facet. In the basic plot above, notice that they differ slightly. This is not ideal for visual comparisons.
3. Modify a few elements related to the theme, including the legend and facet panels
4. Add x and y axis test

```{r message = FALSE, warning = FALSE, error = FALSE, fig.width=12, fig.height= 8}
ggplot(event_counts, aes(fill = AESEV, y = n, x = reorder(AEBODSYS, n))) +
  geom_bar(position="stack", stat="identity", color = "black") +
  coord_flip() +
  scale_fill_brewer(palette = "Blues") +
  facet_wrap(~ TRTA, scales = "free_x") + 
  facetted_pos_scales(y = list(
    scale_y_reverse(breaks = seq(70,0,-10), limits = c(70,0)),
    scale_y_continuous(breaks = seq(0,70,10), limits = c(0,70)))) +
  theme_bw() + 
  theme(legend.position = "bottom",
        legend.title = element_blank(),
        strip.text = element_text(color = "white"),
        strip.background = element_rect(fill = "royalblue"),
        axis.text = element_text(face="bold")) +
  labs(y = "Number of Subjects", x = "System Organ Class")
```