Sharing some general purpose code for making tornado plots in ggplot2.

This uses a hack where the outcomes are transformed into deviations from the base case outcome, plotted using geom_bar, and the axis is manually relabelled.

First, save the data in this format:

Here is the code:

  library(tidyverse)
  library(dplyr)
  library(ggplot2)

  # Base case outcome
  outcome.base <- 10

  # Order by the difference between the High and Low value
  df <- read.csv('DSA.csv') %>%
    group_by(parameter) %>%
    mutate(diff = sum(abs(outcome - outcome.base))) %>%
    arrange(diff, parameter, direction)

  # Keep the top 8 parameters
  df <- tail(df, 8 * 2)

  # Enforce the ordering in the plot
  df <- df %>%
    mutate(parameter = factor(parameter, levels = unique(df$parameter)))

  # Inspect the minimum and maximum values
  min(df$outcome) # -37
  max(df$outcome) # 53

  # Choose axis labels/breaks/limits based on the range of values
  labels <- seq(-45, 60, by = 10)
  breaks <- labels - outcome.base
  limits <- c(-45, 60) - outcome.base

  ggplot(df, aes(x = parameter, y = outcome - outcome.base, fill = direction)) +
    coord_flip() +
    geom_bar(stat = 'identity', position = 'identity', width = 0.5) +
    geom_hline(yintercept = 0) +
    scale_y_continuous(breaks = breaks, labels = labels, limits = limits) +
    geom_text(data = df %>% filter(direction == 'Low', outcome < outcome.base),
              aes(label = value), vjust = 0.5, hjust = 2) +
    geom_text(data = df %>% filter(direction == 'Low', outcome > outcome.base),
              aes(label = value), vjust = 0.5, hjust = -1) +
    geom_text(data = df %>% filter(direction == 'High', outcome < outcome.base),
              aes(label = value), vjust = 0.5, hjust = 2.5) +
    geom_text(data = df %>% filter(direction == 'High', outcome > outcome.base),
              aes(label = value), vjust = 0.5, hjust = -1.5) +
    xlab('') +
    ylab('') +
    theme_bw() +
    theme(panel.grid.minor = element_blank(),
          text = element_text(size = 14),
          legend.position = 'top',
          legend.title = element_blank(),
          legend.key.spacing.x = unit(0.75, 'cm'))

Here is the result: