SUMMARY

COVID-19 cases in Washington State are observed to have substantially different growth rates depending on the region. The core Seattle metro area and the Greater Puget Sound case growth rate is now lower than other regions of the State, despite total case loads still being higher.

GET DATA

Get data from the NYTimes Github Repo.

countydata_file <- "https://raw.githubusercontent.com/nytimes/covid-19-data/master/us-counties.csv"

county_data <- read_csv(countydata_file) 
## Parsed with column specification:
## cols(
##   date = col_date(format = ""),
##   county = col_character(),
##   state = col_character(),
##   fips = col_character(),
##   cases = col_double(),
##   deaths = col_double()
## )

###Patch data

There’s a faulty data point in the data for Benton County (total cases drops on 2020-03-25. This fix just patches that. Issue has been escalated on the github repo.

Create Metro Data Summaries by County

puget_region <- c("Thurston", "Island", "Kitsap", "Pierce", "King", "Snohomish", "Whatcom", "Skagit" )


state_metro_data <- county_data %>%
  filter(state == "Washington") %>%
  mutate(region = ifelse(county %in% puget_region, "Puget Sound", "Greater Washington")) %>%
  yo

Plots of Cases by Region

Plot data is created by summarizing data by metro area. Daily case rates are also computed adn shown as the gray area plot. (the colored line is essnetially the integral of the grey area).

plot_data <- state_metro_data %>%
  group_by(region, date) %>%
  summarize(cases = sum(cases), deaths = sum(deaths)) %>%
  mutate(daily_cases = cases-lag(cases)) %>%
  mutate(daily_deaths = deaths-lag(deaths)) %>%
  arrange(date)%>%
  ungroup() %>%
  mutate(daily_cases = ifelse(is.na(daily_cases), 0, daily_cases)) %>%
  mutate(daily_deaths = ifelse(is.na(daily_deaths), 0, daily_deaths)) %>%
  filter(cases > 1) %>%
  #filter(deaths >1) %>%
  mutate(t_case = daily_cases/(cases-1))%>%
  #mutate(t_death = daily_deaths/(deaths-1)) %>%
  yo

Compute Representative Doubling Time

This can be done a number of ways. Using the mean data for last few days seems to work well. In this case I compute mulitple measurements to do a rough and ready job of estimating uncertainty. It’s not really kisher staistically, but it gives an idea of the magnitude of variance.

Here is the (noisy) doubling time data computed on a daily basis.


fin