Code to create a monthly panel from the Militarized Interstate Disputes dataset

create_monthly_mids_panel.R

# Code to create a monthly, directed dyadic panel of 
# Militarized Interstate Dispute data, with directional initiation
# uses cshapes to create monthly list of all dyads, then uses data.table
# to transform MIDS to a monthly panel and then merge

# Code by Adam Lauretig, 2019
# website: adamlauretig.github.io

rm(list = ls())
options(stringsAsFactors = FALSE)
seed_to_use <- 216
set.seed(seed_to_use)
library(data.table)
library(parallel)
library(cshapes)
load("master_dyads_1946_1980.rdata") 
# for code to create this, see appendix at bottom
setDT(master_dyads)
# revised MID data from http://svmiller.com/gml-mid-data/
mids <- fread("gml-ddy-2.1.csv")

# gets dispute start days ----
mida <- fread("gml-mida-2.1.csv")
mida <- mida[,.(
  dispnum3, stday, stmon, styear, endday, endmon, endyear, fatality, hostlev)]
mida <- mida[ styear > 1945 & styear < 1981]

# use "sprintf" to make sure all days/months preserve the leading zero 
mida[, stmon := sprintf("%02s", stmon)]
mida[, stday := sprintf("%02s", stday)]
mida[, endmon := sprintf("%02s", endmon)]
mida[, endday := sprintf("%02s", endday)]
mida[, `:=`(endday = ifelse(endday == "-9", "01", endday), 
  stday = ifelse(stday == "-9", "01", stday))]

# working w/mids ----
# year range of interest
mids <- mids[ year > 1945 & year < 1981]
mids[, fatal_mid := as.numeric(!(is.na(fatality)) & fatality > 0) ]
mids_sub <- mids[,.(year, ccode1, ccode2, hostlev, sidea1, sidea2, dispnum)]

# merge dispute data with dyad data ----
mid_data <- merge(
  mids_sub, mida, by.x = c("dispnum"), by.y = c("dispnum3"), all.x = TRUE)
# since we're working w/months, set all start dates to "YYYY-MM-01"
mid_data[, start_date := paste0(styear, "-", stmon, "-", "01")]
mid_data[, end_date := paste0(endyear, "-", endmon, "-", "01")]
# clear out some unneeded columns
mid_data[, `:=`(stday = NULL, stmon = NULL, styear = NULL, endday = NULL, 
  endmon = NULL, endyear = NULL, hostlev.y = NULL)]
# here, we only want to look at valid disputes, we're removing any odd ducks
disputes <- mid_data[ !(is.na(dispnum))]
disputes <- disputes[ !(is.na(fatality)) ]
dispute_nums <- unique(disputes$dispnum)


# take the "from" and "to" columns, and create a monthly panel, and 
# code onsets and initiations. returns a data.table
make_monthly_mid <- function(i, dispute_dt = disputes){
  disp <- dispute_nums[i]
  tmp <- dispute_dt[ dispnum == disp ]
  tmp2 <- tmp[ , list(month = seq(as.Date(start_date[1]), 
    as.Date(end_date[1]), by = "month")), 
    by = .(dispnum, ccode1, ccode2, hostlev.x, sidea1, sidea2, fatality)]
  tmp2[, onset := ifelse(month == min(month), 1, 0)]
  tmp2[, onset_na := ifelse(month == min(month), 1, NA)]
  tmp2[, init := ifelse(onset == 1 & sidea1 == 1, 1, 0)]
  tmp2[, init_na := ifelse(onset == 1 & sidea1 == 1, 1, NA)]

return(tmp2)
  
}

# merging and cleaning mids to dyads
monthly_mid_list <- mclapply(1:length(dispute_nums), make_monthly_mid, 
  dispute_dt = disputes, mc.cores = 8)
monthly_mids <- rbindlist(monthly_mid_list)
master_mids <- merge(master_dyads, monthly_mids, 
  by.x = c("ccode1", "ccode2", "year"), by.y = c("ccode1", "ccode2", "month"), 
  all = TRUE) # since we can have multiple mids per year, set all = TRUE
master_mids[, `:=`(
  hostlev.x = ifelse(is.na(hostlev.x), 0, hostlev.x),
  fatality = ifelse(is.na(fatality), 0, fatality),
  onset = ifelse(is.na(onset), 0, onset),
  onset_na = ifelse(is.na(onset_na) & is.na(dispnum), 0, onset_na),
  init = ifelse(is.na(init), 0, init),
  init_na = ifelse(is.na(init_na) & is.na(dispnum), 0, init_na)
  )]
test <- master_mids[,.N, by = .(ccode1, ccode2, year)]
save(master_mids, file = "master_mids.rdata")


# Appendix ----
# create all dyad months
# this tooks about 20 minutes with 8 cores, as a heads up
years <- seq.Date(from = as.Date("1946-01-01"), to = as.Date("1980-12-31"), by = "month")
dl <- distlist(as.Date("1946-01-01"), type="capdist", useGW=FALSE)
dl2 <- dl[ (dl$ccode1 != dl$ccode2), ]
make_dyads <- function(i){
  dyad_year <- years[i]
  dl <- distlist(dyad_year, type="capdist", useGW=FALSE)
  dl2 <- dl[ (dl$ccode1 != dl$ccode2), ]
  dl2$capdist <- NULL
  dl2$year <- dyad_year
  dl2
}
master_dyads <- do.call(rbind, mclapply(1:length(years), make_dyads, mc.cores = 8))
# save(master_dyads, file = "~/Dropbox/Dissertation_data/master_dyads_1946_1980.rdata")