Plot the total number of COVID-19 cases for each US state

covid19_by_us_state.R

library(tidyverse)
library(ggplot2)
library(shadowtext)

#John J Davis IV
#Biomechanics PhD Student
#Indiana University School of Public Health
#Twitter:  @JDRuns
#Email:    jjd1@iu.edu
#25 March 2020

#NOTE: This still breaks Lat/Long for some instances. 
#It's an easy fix with an inner_join()

#This code is public domain

#Adapted in part from code by John Burn-Murdoch

#Read data directly from Johns Hopkins GitHub
covid_df <- read_csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_19-covid-Confirmed.csv")

#Gather the state-wide data, ignoring cruise ships
covid_state_data <- covid_df %>% filter(`Country/Region` == "US") %>%
  filter(!grepl(".,.", `Province/State`)) %>%
  filter(!`Province/State` %in% c("Diamond Princess", 
                                  "Grand Princess", 
                                  "US", "United States Virgin Islands")) %>%
  rename(state = `Province/State`)  %>% 
  pivot_longer(-(1:4), names_to = "date", values_to = "cases")

#Gather county-wide data (collected up until 3/9) and pool with state-data
covid_us <- covid_df %>% filter(`Country/Region` == "US") %>%
  filter(grepl(paste(".",
                     paste(state.abb, collapse="|"), 
                     ".", sep = ""),
               `Province/State`)) %>%
  separate(`Province/State`, into = c("city", "state"), sep = ",\\s") %>%
  group_by(state) %>% summarise_if(is.numeric, sum) %>%
  mutate(`Country/Region` = "US") %>%
  select(state, `Country/Region`, everything()) %>%
  mutate(state = state.name[match(state, state.abb)]) %>% 
  pivot_longer(-(1:4), names_to = "date", values_to = "cases") %>%
  bind_rows(covid_state_data) %>%
  mutate(date = as.Date(date, "%m/%d/%y"))  %>%
  arrange(state, date)

#Plot total cases as a function of time, on log scale.
display_day <- format(Sys.Date() + 7, format="%m/%d/%y")

state_terr_abb <- c(state.abb, "DC", "VI", "GU", "PR")
state_terr_name <- c(state.name, "District of Columbia", "Virgin Islands", "Guam", "Puerto Rico")

covid_us %>% filter(cases > 10) %>%
  ggplot(aes(x=date, y=cases, color = state)) + 
  geom_line(lwd=1) +
  geom_point(size=2, pch = 21, fill="white", alpha=0.8) +
  geom_shadowtext(data = . %>% 
                    filter (cases > 1) %>%
                    top_n(1, date), 
                  aes(label = state_terr_abb[match(state, state_terr_name)]),
                  bg.color = "white",
                  size=5,
                  position = position_jitter(height = 0.05, width = 0.8)
                  ) +
  scale_y_log10(limits = c(10, 50000), breaks = c(10,100,1000,10000,50000)) + 
  scale_x_date(limits = as.Date(c("2/25/2020", display_day), "%m/%d/%y"),
               date_breaks = "1 month",
               date_labels = "%b %d"
  )  +
  labs(x = "Date", subtitle = "Total cases (after 10th case) by US state") + 
  theme(legend.position = "none",
        text = element_text(size=16),
        strip.text = element_text(size=12))