R/create_leaderdays.R

In peacesciencer: Tools and Data for Quantitative Peace Science Research

#' Create leader-days from leader data
#'
#' @description \code{create_leaderdays()} allows you to generate leader-day data from
#' leader-level data provided in \pkg{peacesciencer}.
#'
#' @return \code{create_leaderdays()} takes leader-level data available in \pkg{peacesciencer}
#' and returns a leader-day-level data frame.
#'
#' @details \code{create_leaderdays()}, as of writing, only supports the Archigos data set of leaders. I envision this function
#' being mostly for internal uses. Basically, \code{create_leaderyears()} effectively starts by first running a version of
#' \code{create_leaderdays()}. So, why not have this function too?
#'
#' @author Steven V. Miller
#'
#' @references
#'
#' Goemans, Henk E., Kristian Skrede Gleditsch, and Giacomo Chiozza. 2009. "Introducing Archigos: A Dataset of Political Leaders"
#' \emph{Journal of Peace Research} 46(2): 269--83.
#'
#' @param system a leader system with which to create leader-days. Right now, only "archigos" is supported.
#' @param standardize a character vector of length one: "cow", "gw", or "none". If "cow", the function standardizes the leader-days to
#' just those that overlap  with state system membership in the Correlates of War state system (see: \code{cow_states}). If "gw", the function
#' standardizes the leader-days to just those that overlap with the state system dates of the Gleditsch-Ward date (see: \code{gw_states}). If
#' "none", the function returns all leader-days as presented in Archigos (which is nominally denominated in Gleditsch-Ward state system codes,
#' if not necessarily Gleditsch-Ward state system dates). Default is "none".
#'
#' @examples
#' \donttest{
#' create_leaderdays()
#'
#' create_leaderdays(standardize = "gw")
#' }
#'
create_leaderdays <- function(system = "archigos", standardize = "none") {

  if (!(standardize %in% c("none", "cow", "gw")) | length(standardize) > 1) {

    stop("The `standardize` argument must be either 'none', 'cow', or 'gw'. Pick just one.")
  }

  if (system == "archigos") {

    archigos %>%
      rowwise() %>%
      mutate(date = list(seq(.data$startdate, .data$enddate, by="1 day"))) %>%
      unnest(.data$date) %>%
      mutate(year = .pshf_year(.data$date),
             yrinoffice = (.data$year - .pshf_year(.data$startdate)) + 1) -> leaderdays

    if (standardize == "gw") {

      gw_statedays <- create_statedays(system = 'gw')

      leaderdays %>%
        semi_join(., gw_statedays) -> data

      attr(data, "ps_data_type") = "leader_day"
      attr(data, "ps_system") = "gw"

      data <- subset(data, select=c("obsid", "gwcode", "leader", "date", "yrinoffice"))

    } else if (standardize == "cow") {

      cow_statedays <- create_statedays()


      leaderdays %>%
        left_join(., gw_cow_years %>% select(.data$gwcode, .data$ccode, .data$year)) %>%
        select(.data$obsid, .data$gwcode, .data$ccode, everything()) -> leaderdays

      # I need to do some ad hoc corrections here
      # There are pieces of the gw_cow_years data frame where G-W says the state doesn't exist but it appears in the CoW data.
      # Because G-W is the "master" here, this is going to happen.
      # The gw_cow_years data would be functionally correct here, but that's immaterial for matching it to the Archigos data.
      # gw_cow_years %>% filter(is.na(gwcode) & year >= 1870) %>% group_by(stateabb) %>% summarize(ccode = max(ccode, na.rm=T), n = n(), min = min(year), max = max(year))
      # For example: G-W have Morocco drop out at 1904. CoW has Morocco drop out at 1912.
      # If you want to standardize to CoW, you're going to miss that.
      # You'll also miss that G-W has Saudi Arabia start in 1932, but CoW has it at 1927.
      # Archigos has leader data for the Saudis to 1927, so you'll want that.
      # Additionally,  we need to get more granular with the transitions in 1990 for Yemen and Germany.
      # Saleh al-hashidi and Kohl presided over both transitions, though CoW sees the implications differently for the state system

      leaderdays %>%
        mutate(ccode = case_when(
          # A few ad hoc corrections here. First CoW, has these state periods for Morocco and Saudi Arabia where G-W does not
          .data$gwcode == 600 & (.data$year >= 1905 & .data$year <= 1912) ~ 600,
          .data$gwcode == 670 & (.data$year >= 1927 & .data$year <= 1932) ~ 670,
          # Next, we need to get more granular with the transitions in 1990 for Yemen and Germany
          # Saleh al-hashidi and Kohl presided over both transitions, though CoW sees the implications differently for the state system
          .data$obsid == "YEM-1978" & .data$date >= "1990-05-22" ~ 679,
          .data$obsid == "YEM-1978" & .data$date < "1990-05-22" ~ 678,
          .data$obsid == "GFR-1982" & .data$date >= "1990-10-03" ~ 255,
          .data$obsid == "GFR-1982" & .data$date < "1990-10-03" ~ 260,
          TRUE ~ .data$ccode
        )) -> leaderdays

      leaderdays %>%
        semi_join(., cow_statedays) -> data

      attr(data, "ps_data_type") = "leader_day"
      attr(data, "ps_system") = "cow"

      data <- subset(data, select=c("obsid", "ccode", "leader", "date", "yrinoffice"))

    } else { # assuming standardize = "none"

      leaderdays -> data

      attr(data, "ps_data_type") = "leader_day"
      attr(data, "ps_system") = "gw"

      data <- subset(data, select=c("obsid", "gwcode", "leader", "date", "yrinoffice"))

    }


  } else if(system != "archigos") {

    stop("Right now, only the Archigos leader data are supported.")

  }

  return(data)
}