Explores Covid-19 data truncation in England (i.e when data is updated in later releases) using an experimental model in EpiNow2. By default looks at test positive cases but this can be updated by changing the selected variable in line 27.

explore-covid-19-data-truncation

# Note: estimate_truncation is experimental so use this for exploratory purposes
# only and/or with a high level of interpretation
# Packages ----------------------------------------------------------------
# install packages
# install.packages(c("data.table", "purrr", "remotes", "EpiNow2"))
library(data.table)
library(purrr)
library(covidregionaldata)
library(EpiNow2)

# Set cores ---------------------------------------------------------------
options(mc.cores = 4)

# Download data -----------------------------------------------------------
end_date <- Sys.Date()
uk_df <- map(seq(end_date - 13, end_date, by = "day"),
             ~ suppressMessages(get_regional_data("uk", 
                                                  nhsregions = TRUE,
                                                  release_date = .)))

# Convert to expected variables for modelling -----------------------------
process_df <- function(df) {
  df <- data.table::as.data.table(df)
  df <- df[region %in% "England"]
  df <- df[, .(date, confirm = as.integer(cases_new))]
  df <- df[!is.na(confirm)]
  df <- df[date >= (Sys.Date() - 4 * 7)]
}
model_df <- map(uk_df, process_df)

# Look at the data and prune out obvious truncation -----------------------
map(model_df, tail)
# get rid of NA data
model_df <- map(model_df, ~ .[!is.na(confirm)])
# get rid of last value if always zero
model_df <- map(model_df, ~ .[date < max(date)])

# Fit truncation model ----------------------------------------------------
#see ?estimate_truncation for details
truncation <- estimate_truncation(model_df, max_truncation = 8)

# Evaluate truncation -----------------------------------------------------
plot(truncation)