COVID/COVID_new.R

starttime <- Sys.time()
#setup ----
library(tidyverse)
library(lubridate)
library(RcppRoll)
library(ggrepel)

setwd("~/Documents/dataProjects/COVID")

if(file.exists("data_download_time.Rda")) {
  load("data_download_time.Rda")
} else {
  downloaded_dttm <- ymd_hms(paste(max(read_csv(paste0('data/us-counties-', year(Sys.Date()),'.csv'))$date), '00:00:00'))
}

# load data and download new data if needed ----

us_county_data <- list(NULL)
for(year in 2020:year(Sys.Date())){
  if(year(downloaded_dttm)>year){
    us_county_data[[year]] <- read_csv(paste0('data/us-counties-',year,'.csv'))
  } else {
    if(as.double(difftime(downloaded_dttm, Sys.time()), units = "hours") < -1){
      print("Downloading today's data")
      download.file(url = paste0('https://github.com/nytimes/covid-19-data/raw/master/rolling-averages/us-counties-',year,'.csv'),
                    destfile = paste0('data/us-counties-',year,'.csv'))
      us_county_data[[year]] <- read_csv(paste0('data/us-counties-',year,'.csv'))
    } else {
      print("data is current")
      us_county_data[[year]] <- read_csv(paste0('data/us-counties-',year,'.csv'))
    }
  }
}
downloaded_dttm <- Sys.time()
save(downloaded_dttm, file = "data_download_time.Rda")
us_county_data <- bind_rows(us_county_data)
maxdate <- max(us_county_data$date)

#load population data
us_county_pop <- us_county_pop <- read.csv("data/co-est2020-alldata.csv")
us_county_pop['geoid'] <- paste0('USA-',strrep('0', 2 - nchar(us_county_pop$STATE)), us_county_pop$STATE, strrep('0', 3 - nchar(us_county_pop$COUNTY)), us_county_pop$COUNTY)
us_county_data['population'] <- us_county_pop[match(us_county_data$geoid,us_county_pop$geoid), 20]

#get cumulative two week totals
us_county_data <- us_county_data %>%
  complete(date = full_seq(date, period = 1), fill = list(cases = 0)) %>%
  group_by(geoid) %>%
  arrange(date) %>%
  mutate(twowkcases = roll_sumr(cases, n = 7, na.rm = TRUE)/7) %>%
  mutate(active_estimate = twowkcases/population * 100000)

#important counties
us_county_recent <- us_county_data %>% filter(date %in% maxdate)
important_counties <- bind_rows(
  data.frame(state = 'Washington',
             county = c("Clallam")),
  data.frame(state = "Wisconsin",
             county = c("Milwaukee",
                        "Dane")),
  data.frame(state = "Illinois",
             county = c("Lake")),  
  data.frame(state = "Utah",
             county = c("Salt Lake")))
important_counties['geoid'] <- us_county_recent[match(paste(important_counties$county,important_counties$state), paste(us_county_recent$county,us_county_recent$state)), 2]

nearby_counties <- bind_rows(
  data.frame(state = 'Washington',
             county = c("Clallam",
                        'Kitsap',
                        'King',
                        'Jefferson')))
nearby_counties['geoid'] <- us_county_recent[match(paste(nearby_counties$county,nearby_counties$state), paste(us_county_recent$county,us_county_recent$state)), 2]

# make charts ----
ggplot(data = us_county_data %>% filter(county == "Clallam")) +
  geom_line(aes(x = date,
                y = cases_avg_per_100k),
                color = 'blue',
            size = 0.5) +
  geom_line(aes(x = date,
                y = active_estimate),
            color = 'red',
            size = 0.5) +
  scale_x_date(date_breaks = "1 month", date_labels = "%b\n%Y", minor_breaks = "1 week", expand = expansion(mult = c(0.01, 0.07))) +
  scale_y_continuous(expand = expansion(mult = c(0,0.1))) +
  labs(title = "Case rate - nearby counties",
       subtitle = paste("Through", format(maxdate, "%B %e, %Y")),
       x = "Date",
       color = NULL,
       y = "7-day trailing average new cases per day per 100,000 people",
       caption = "data from The New York Times") +
  theme_bw() +
  theme(panel.grid.major.x = element_line(colour="black", size = 0.1), 
        axis.text.x = element_text(angle = 60, hjust = 1),
        plot.subtitle = element_text(color = "grey50"),
        plot.caption = element_text(color = "grey50"),
        legend.position = "bottom") +
  coord_cartesian(ylim = c(0,NA),
                  clip = "off")


ggplot() +
  geom_line(data = us_county_data %>% 
              filter(geoid %in% important_counties$geoid),
            aes(x = date,
                y = cases_avg_per_100k,
                color = county)) +
  geom_label_repel(data = us_county_data %>% 
                     filter(geoid %in% important_counties$geoid,
                            county != c("Jefferson, King, Kitsap"),
                            date %in% maxdate),
                   aes(x = date + 0.5,
                       y = cases_avg_per_100k,
                       label = paste0(county,' - ', round(cases_avg_per_100k,0)),
                       fill = county),
                   hjust = "outward",
                   direction = "y",
                   size = 2.5,
                   nudge_x = 4,
                   box.padding = 0.01,
                   min.segment.length = 0,
                   segment.color = "black",
                   segment.size = 0.1) +
  scale_color_brewer(palette = "Set2", guide = "none") +
  scale_fill_brewer(palette = "Set2", guide = "none") +
  scale_x_date(date_breaks = "1 month", 
               date_labels = "%b\n%Y", 
               minor_breaks = "1 week", 
               expand = expansion(mult = c(0.01, .07))) +
  scale_y_continuous(expand = expansion(mult = c(0,0.1))) +
  labs(title = "Active case estimate",
       subtitle = paste("Through", format(maxdate, "%B %e")),
       x = "Date",
       y = "7-day trailing average new cases per day per 100,000 people",
       color = "County",
       caption = "data from The New York Times") +
  theme_bw() +
  theme(panel.grid.major.x = element_line(colour="black", size = 0.1), 
        axis.text.x = element_text(angle = 60, hjust = 1),
        plot.subtitle = element_text(color = "grey50"),
        plot.caption = element_text(color = "grey50")) +
  coord_cartesian(ylim = c(0,NA))

message(difftime(Sys.time(), starttime))
rm(starttime)