2020-03-19-new-cases.Rmd

---
title: "New Cases"
description: |
  New Cases. Written 3/18, 4:20pm.
site: distill::distill_website
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE)
install.load::install_load(c('distill','dplyr', 'tidyverse', 'ggplot2', 'plotly', 'rmarkdown', 'magrittr', 'RCurl', 'lubridate', 'DT', 'ggthemes', 'zoo', 'rvest', 'patchwork'))
ggplot2::theme_set(theme_minimal())

read_data <- function(type='Deaths') {
  
  fp_data = paste0('https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_19-covid-', type ,'.csv')
  
  # Read in data
  df = read_csv(fp_data) %>%
    select(-Lat, -Long) %>%
    gather(date_raw, stat, 
           -`Province/State`, -`Country/Region`) %>%
    mutate(date = mdy(date_raw))
  
  df[[type]] = df$stat
  
  # Clean column names
  colnames(df) %<>% 
    tolower() %>% 
    str_replace_all(fixed('/'), '_') %>%
    str_replace_all(fixed(' '), '_')
  
  # Re-label states
  states = tibble(state_abb = state.abb, province_state=state.name)
  # Join states together
  df %<>%
    left_join(states, by='province_state') %>%
    mutate(us_state = str_extract(province_state, '[A-Z][A-Z]$'),
           us_state = ifelse(!is.na(us_state), us_state, state_abb))
  
  # Assert us_state is NA for non-US country_region
  stopifnot(is.na(df %>% filter(country_region != 'US') %>% pull(us_state) %>% unique()))
  
  df %>% 
    select(-stat, -date_raw, -state_abb) %>%
    select(country_region, province_state, us_state, date, everything()) %>%
    arrange(country_region, province_state, us_state, date) %>%
    return()
}

get_df_population <- function(){
  df_html = xml2::read_html('http://www.worldometers.info/world-population/population-by-country/') %>%
    html_nodes('table') %>%
    html_table() 
  df_pop = df_html[[1]]
  
  colnames(df_pop) %<>% 
    tolower() %>% 
    str_replace_all(' ', '_') %>%
    str_replace_all('\\(', '_') %>%
    str_replace_all('\\)', '_') %>%
    str_replace_all(fixed('__'), '_') %>%
    str_replace_all('\\_$', '') 
  
  df_pop %>%
    select(country_or_dependency, population_2020) %>%
    rename(pop = population_2020,
           country_region = country_or_dependency) %>%
    mutate(country_region = ifelse(country_region == 'United States', 'US', country_region),
           pop = readr::parse_number(pop))
}



dfd <- read_data(type='Deaths')
dfc <- read_data(type='Confirmed')
dfr <- read_data(type='Recovered')

# join all dataframes together with population
df <- dfd %>% 
  left_join(dfc) %>% 
  left_join(dfr) %>%
  left_join(get_df_population())

```

```{r}
countries = c('US', 'Spain', 'Italy', 'Germany', 'Iran', 'China')

df_new = df %>%
  filter(country_region %in% countries) %>%
  group_by(country_region, date) %>%
  summarize(deaths = sum(deaths),
            confirmed = sum(confirmed),
            recovered = sum(recovered),
            pop = max(pop)) %>%
  mutate(
    deaths_pop = deaths/pop,
    confirmed_pop = confirmed/pop,
    cfr = deaths/confirmed,
    cfr2 = deaths/(deaths + recovered),
    new_deaths = deaths - lag(deaths),
    new_confirmed = confirmed - lag(confirmed),
    new_recovered = recovered - lag(recovered),
    deaths_ma = zoo::rollmean(new_deaths, k=5, fill=NA),
    confirmed_ma = zoo::rollmean(new_confirmed, k=5, fill=NA),
    deaths_ma_pop = zoo::rollmean(new_deaths/pop, k=5, fill=NA),
    confirmed_ma_pop = zoo::rollmean(new_confirmed/pop, k=5, fill=NA),
    confirmed_2daypct = (new_confirmed - lag(new_confirmed, 2))/lag(new_confirmed, 2)
  ) 

#plot_line(v = 'confirmed_2daypct', 'hi', 'lag2 confirmed')
```

## New Cases, Deaths, and CFR

### Confirmed
```{r}
df_new %>%
  ggplot(aes(x = date, y = new_confirmed, color = country_region)) +
  geom_line() +
  scale_color_pander(name='Country') + 
  labs(title = 'Daily Confirmed Cases by Country',
       y = 'Deaths',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```

### Deaths
```{r}
df_new %>%
  ggplot(aes(x = date, y = new_deaths, color = country_region)) +
  geom_line() +
  scale_color_pander(name='Country') + 
  theme(legend.position = 'top') +
  labs(title = 'Daily Deaths by Country',
       y = 'Deaths',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```

### Deaths (Scaled by Population)

```{r}
df_new %>%
  ggplot(aes(x = date, y = deaths_pop, color = country_region)) +
  geom_line() +
  scale_y_continuous(labels = scales::percent) + 
  scale_color_pander(name='Country') + 
  theme(legend.position = 'top') +
  labs(title = 'Daily Deaths by Country, Scaled by Population',
       y = 'Deaths/Population Size',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```

### Recovered

```{r}
df_new %>%
  ggplot(aes(x = date, y = new_recovered, color = country_region)) +
  geom_line() +
  scale_color_pander(name='Country') + 
  theme(legend.position = 'top') +
  labs(title = 'Daily Recovered by Country',
       y = 'Recovered',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```



## Cumulative Confirmed, Deaths, Recovered

```{r cumulative, results='asis'}

plot_line <- function(v, title, ytitle, ...) {
  
  p <- df_new %>%
    ggplot(aes_string(x = 'date', y = v, color = 'country_region')) +
    geom_line() +
    scale_color_pander(name='Country') + 
    theme(legend.position = 'top') +
    labs(title = title,
         y = ytitle,
         x = '',
         caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19',
         ...)
  if (str_detect(v, 'pop')){
    p <- p + scale_y_continuous(labels = scales::percent)
  }
  return(p)
}
```

## New Confirmed Cases, Deaths, and Recovered

```{r}
p1 = plot_line(v='new_confirmed', 
          title = 'Daily Confirmed Cases',
          ytitle = '')
p2 = plot_line(v='new_deaths', 
          title = 'Daily Deaths',
          ytitle = '')
p3 = plot_line(v='new_recovered', 
          title = 'Daily Recovered',
          ytitle = '')

p1 
p2
p3
```

## Cumulative Confirmed Cases, Deaths, and Recovered

```{r, "l-screen-inset shaded"}
p1 = plot_line(v='confirmed', 
          title = 'Confirmed',
          ytitle = '')
p2 = plot_line(v='deaths', 
          title = 'Deaths',
          ytitle = '')
p3 = plot_line(v='recovered', 
          title = 'Recovered',
          ytitle = '')

combined <- p1 + p2 + p3 & theme(legend.position = "bottom") 
combined + plot_layout(guides = "collect") + 
  plot_annotation(
    title='Cumulative Confirmed Cases, Deaths, and Recovered',
    caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19'
  )
```

```{r}
plot_line(v='deaths', 
          title = 'Cumulative Daily Deaths',
          ytitle = 'Cumulative Deaths')
```


```{r}
plot_line(v='confirmed', 
          title = 'Cumulative Daily Confirmed Cases',
          ytitle = 'Confirmed Cases',
          subtitle = 'Moving Average = (5-day sum)/5')
```


## Preliminary Case Fatality Rate
True case fatality rate is unreliable so early in a pandemic because 

1. Testing is frequently limited to those who are most ill
2. Testing is not widely available
3. Reporting is not widely reliable

### CFR: Deaths/Confirmed
```{r}
df_new %>%
  ggplot(aes(x = date, y = cfr, color = country_region)) +
  geom_line() +
  theme(legend.position = 'top') +
  scale_y_continuous(labels = scales::percent, limits = c(0, 0.1)) + 
  scale_color_pander(name='Country') + 
  labs(title = 'Preliminary* Case Fatality Rate by Country as of a Given Date',
       subtitle = 'Where CFR = cumulative deaths/cumulative confirmed',
       y = 'CFR',
       x = '',
       caption = '\n*CFR is known to be unreliable so early in a pandemic\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```

### CFR: Deaths/(Deaths + Recovered)

```{r}
df_new %>%
  filter(recovered > 0) %>%
  ggplot(aes(x = date, y = cfr2, color = country_region)) +
  geom_line() +
  theme(legend.position = 'top') +
  scale_y_continuous(labels = scales::percent) + 
  scale_color_pander(name='Country') + 
  labs(title = 'Preliminary* Case Fatality Rate by Country as of a Given Date',
       subtitle = 'Where CFR = cumulative deaths/(cum. deaths + cum. recovered)',
       y = 'CFR',
       x = '',
       caption = '\n*CFR is known to be unreliable so early in a pandemic\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```


## Moving Averages

### Confirmed
```{r}
df_new %>%
  ggplot(aes(x = date, y = confirmed_ma, color = country_region)) +
  geom_line() +
  scale_color_pander(name='Country') + 
  labs(title = 'Moving Average of Confirmed Cases',
       subtitle = '*MA = (sum of prior 5 days) / 5',
       y = ' MA of New Confirmed Cases*',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```

### Confirmed (Scaled by Population)

```{r}
df_new %>%
  ggplot(aes(x = date, y = confirmed_ma_pop, color = country_region)) +
  geom_line() +
  scale_color_pander(name='Country') + 
  scale_y_continuous(labels = scales::percent) + 
  labs(title = 'Moving Average of Confirmed Cases, Scaled by Population',
       subtitle = '*MA = (sum of prior 5 days) / 5',
       y = ' MA of New Confirmed Cases*',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```

### Deaths

```{r}
df_new %>%
  ggplot(aes(x = date, y = confirmed_ma, color = country_region)) +
  geom_line() +
  scale_color_pander(name='Country') + 
  labs(title = 'Moving Average of Deaths',
       subtitle = '*MA = (sum of prior 5 days) / 5',
       y = ' MA of Deaths*',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```

### Deaths (Scaled by Population)

```{r}
df_new %>%
  ggplot(aes(x = date, y = deaths_ma_pop, color = country_region)) +
  geom_line() +
  scale_color_pander(name='Country') + 
  scale_y_continuous(labels = scales::percent) + 
  labs(title = 'Moving Average of Deaths, Scaled by Population',
       subtitle = '*MA = (sum of prior 5 days) / 5',
       y = ' MA of Deaths*',
       x = '',
       caption = '\nData Source: github.com/CSSEGISandData/COVID-19\nAnalysis: github.com/bryanwhiting/covid19')
```