In naturewillconfess/rucovidr: Scrapes COVID-19 data from the Rospotrebnadzor website

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(rucovidr)
library(dplyr)
library(zoo)
library(ggplot2)
library(tidyr)
library(lubridate)
library(xml2)
data(old_cases, package = "rucovidr")

Cases and testing in Russia

Let's first scrape and wrangle the necessary data - regional data on new cases and federal data on the tests:

news <- get_news()
tests_fed <- get_tests_federal(news)
cases <- get_cases(news)
cases <- rbind(old_cases, cases)

tests_fed <- tests_fed[-c(1:(which(!is.na(tests_fed$tests))[1]-1)),]
tests_fed$tests <-  round(na.approx(tests_fed$tests))
tests_fed$tests <-  c(NA, tests_fed$tests[-1] - tests_fed$tests[-length(tests_fed$tests)])
###correcting apparent errors in the data, specifically zero and negative values. First two weeks are so messy it's not even worth analyzing imo
tests_fed <- tests_fed[-c(1:11),]
tests_fed$tests[tests_fed$date %in% ymd(c("2020-03-13", "2020-03-14"))] <- sum(tests_fed$tests[tests_fed$date %in% ymd(c("2020-03-13", "2020-03-14"))])/2
tests_fed$tests[tests_fed$date %in% ymd(c("2020-03-31", "2020-04-01", "2020-04-02"))] <- sum(tests_fed$tests[tests_fed$date %in% ymd(c("2020-03-31", "2020-04-01", "2020-04-02"))])/3
tests_fed$tests[tests_fed$date %in% ymd(c("2020-08-22", "2020-08-23"))] <- sum(tests_fed$tests[tests_fed$date %in% ymd(c("2020-08-22", "2020-08-23"))])/2


df_fed <- 
  cases %>%
  group_by(date) %>%
  summarize(cases = sum(as.numeric(cases), na.rm = TRUE), .groups = "drop") %>%
  full_join(tests_fed) %>%
  mutate(cases = replace_na(cases, 0),
         positivity = cases/tests) %>%
  arrange(date) 

ggplot(df_fed, aes(date, cases)) + geom_smooth() + geom_point(alpha = 0.2, colour = "red")
ggplot(df_fed, aes(date, tests)) + geom_smooth() + geom_point(alpha = 0.2, colour = "red")
ggplot(df_fed, aes(date, positivity)) + geom_smooth() + geom_point(alpha = 0.2, colour = "red")

Cases and testing in Moscow

tests_reg <- get_tests_regional(news)
tr1 <- tests_reg
tests_reg <- tr1

tests_moscow <- 
  tests_reg %>%
  complete(nesting(region, date)) %>%
  filter(region == "Москва") %>%
  mutate(abs = c(percapita[1], abs[-1]),
         abs = c(NA, abs[-1] - abs[-length(abs)])) %>% ### a mistake in the first regional report
  select(-percapita)

df_moscow <-
  cases %>%
  filter(region == "Москва") %>%
  full_join(tests_moscow) %>%
  arrange(date) %>%
  mutate(cases = as.numeric(cases),
         cases_rollsum = c(rep(0, 6), rollsum(cases, 7))) %>%
  filter(!is.na(abs)) %>%
  select(-cases) %>%
  rename(cases = cases_rollsum,
         tests = abs) %>%
  mutate(positivity = cases/tests)

ggplot(df_moscow, aes(date, cases)) + geom_smooth() + geom_point(alpha = 0.2, colour = "red")
ggplot(df_moscow, aes(date, tests)) + geom_smooth() + geom_point(alpha = 0.2, colour = "red")
ggplot(df_moscow, aes(date, positivity)) + geom_smooth() + geom_point(alpha = 0.2, colour = "red")

naturewillconfess/rucovidr documentation built on Jan. 1, 2021, 11:33 a.m.