get R estimates for all times

get_covid19_nowcasts_all_times.r

##' Get estimates shown at \url{https://epiforecasts.io/covid} for all times
##'
##' Nowcasts at \url{https://epiforecasts.covid} and the related github repository at
##' \url{https://github.com/epiforecasts/covid-rt-estimates} only cover the last 3
##' months. This function downloads all available estimates and applies a median
##' average to the provided quantiles to provide an estimate of a time series
##' covering all times available
##'
##' @param dataset character; data set corresponding to directories at
##' \url{https://github.com/epiforecasts/covid-rt-estimates}. Default is
##' \code{"national/cases"}, other examples of possible values are:
##' \code{"subnational/brazil/cases"} or
##' \code{"subnational/united-kingdom/admission"}
##' @param variable character; quantity to consider; by default: "rt", but
##' could also be, e.g.. "growth_rate", or "cases_by_infection"; corresponds to
##' the name of \code{.csv} files in the summary directories at
##' \url{https://github.com/epiforecasts/covid-rt-estimates}
##' @param earliest_date character; the earliest date to consider; by default:
##' "30 August 2020" (the day after estimates started being produced using
##' EpiNow2). It is not recommended to set this to an earlier date as it would
##' lead to a mixture of methods.
##' @return a list of data frames containing estimates for all times
##' @importFrom gh gh
##' @importFrom readr read_csv
##' @importFrom dplyr bind_rows group_by summarise
##' @importFrom tidyr pivot_longer pivot_wider
##' @author Sebastian Funk
get_covid19_nowcasts_all_times <-
  function(dataset = "national/cases", variable = "rt",
           earliest_date = "2020-08-30") {
  owner <- "epiforecasts"
  repo <- "covid-rt-estimates"
  path <- paste(dataset, "summary", paste0(variable, ".csv"), sep = "/")

  query <- "/repos/{owner}/{repo}/commits?path={path}"

  if (!is.null(earliest_date)) {
    query <- paste0(query, "&since={date}")
  }

  commits <-
    gh::gh(query,
           owner = owner,
           repo = repo,
           path = path,
           date = earliest_date,
           .limit = Inf)

  shas <- vapply(commits, "[[", "", "sha")

  estimates <-
    lapply(
      shas,
      function(sha)
        readr::read_csv(
                 paste("https://raw.githubusercontent.com", owner, repo,
                       sha, path, sep = "/"))
    )

  median_estimates <- estimates %>%
    bind_rows() %>%
    filter(type == "estimate") %>%
    select(-strat, -type, -mean, -sd) %>%
    pivot_longer(c(median, starts_with("lower_"), starts_with("upper_"))) %>%
    group_by_at(vars(1, "date", "name")) %>%
    summarise(value = median(value), .groups = "drop") %>%
    pivot_wider()

  return(median_estimates)
}

library("dplyr")
library("readr")
library("tidyr")
library("gh")

rt_brazil <- get_covid19_nowcasts_all_times("subnational/brazil/cases")