Q_rolling

labels = read.csv("breaks.csv")

Smooth

Big Corpus – Size 1000, step 200, with punct

# Prepare data
dists = read.csv("FINAL_PREDICTIONS_s1000-step200-nopunct_BigData.csv")
# Get average point
x = lapply(strsplit(sub(pattern = "^[^_]+\\_", "", dists[,2]), "-"), as.numeric)
x = lapply(x, mean)
#labels = read.csv("subdivisions.csv")

#Plot
couls = matrix(ncol = 1, nrow = ncol(dists)-3, data=c(
  "dodgerblue2", "khaki2", 
  "green4",
  "#6A3D9A", # purple
  "#FF7F00", # orange
  "gold1", "palegreen2", # red
  "skyblue2", "#FB9A99", # lt pink
  "#E31A1C",
  "#CAB2D6", # lt purple
  "black",
  "#FDBF6F", # lt orange
  "gray70", 
  "maroon", "orchid1", "deeppink1", "blue1", "steelblue4",
  "darkturquoise", "green1", "yellow4", "yellow3",
  "darkorange4", "brown"
)[1:(ncol(dists)-3)], dimnames = list(colnames(dists)[4:ncol(dists)], "color"))

for(i in 4:ncol(dists)){
  if(i == 4){
    plot(x, smooth(dists[,4], kind = "3RSR"), type="l", ylim = c(round(min(dists[, 4:ncol(dists)]), digits = 1),1), lty=1, ylab = "decision function", xlab = "words", col=couls[colnames(dists)[i],], ljoin=3)
  }
  else{
    lines(x, smooth(dists[,i], kind = "3RSR"), col=couls[colnames(dists)[i],], ljoin=3)
  }
}
legend(1,1, legend=colnames(dists)[4:ncol(dists)], col=couls[colnames(dists[4:ncol(dists)]),], lty=1, ncol=1, cex = 0.5)
abline(v=labels[,2], col="grey", lty=2)
text(x = labels[,2], y = 0.5,  labels[,1], srt=90, cex=0.8)

Clean corpus – Size 1000, step 200, with punct

# Prepare data
dists = read.csv("FINAL_PREDICTIONS_s1000-step200-nopunct.csv")
# Get average point
x = lapply(strsplit(sub(pattern = "^[^_]+\\_", "", dists[,2]), "-"), as.numeric)
x = lapply(x, mean)
#labels = read.csv("subdivisions.csv")

#Plot
#dists2[, 
#dists = dists[, !colnames(dists) %in% c("ericTrump", "melaniaTrump", "GeneralFlynn")]
for(i in 4:ncol(dists)){
  if(i == 4){
    plot(x, smooth(dists[,4], kind = "3RSR"), type="l", ylim = c(round(min(dists[, 4:ncol(dists)]), digits = 1),1), lty=1, ylab = "decision function", xlab = "words", col=couls[colnames(dists)[i],], ljoin=3)
  }
  else{
    lines(x, smooth(dists[,i], kind = "3RSR"), col=couls[colnames(dists)[i],], ljoin=3)
  }
}
legend(1,1, legend=colnames(dists)[4:ncol(dists)], col=couls[colnames(dists[4:ncol(dists)]),], lty=1, ncol=1, cex=0.8)
abline(v=labels[,2], col="grey", lty=2)
text(x = labels[,2], y = 0.5,  labels[,1], srt=90, cex=0.8)