cdaten = read.table(file.path(pfad, "compdaten.txt"), header=T)
vpn = cdaten$Vpn
vpn[vpn=="Sfr"] = "sfr"
vpn = factor(vpn)
cdaten[,5] = vpn
tp = substring(cdaten$Satz, 1, 1)
cdaten = cbind(cdaten, factor(tp))
names(cdaten)[6] = "Wtype"
n = paste("s", substring(cdaten$Satz, 2, 2), sep="")
cdaten = cbind(cdaten, factor(n))
names(cdaten)[7] = "Stype"
cdaten = cdaten[,-4]
# vowel duration in /Ik/ vs /Iks/
# i.e. s2 vs s3
temp = cdaten$Stype %in% c("s2", "s3")
ca = cdaten[temp,]
cb = cdaten[!temp,]
ca$Stype = factor(ca$Stype)
cb$Stype = factor(cb$Stype)

# compute difference values
temp = ca$Stype == "s2"
x = ca[temp,]
y = ca[!temp,]
any(paste(x$Vpn, x$Wtype) !=paste(y$Vpn, y$Wtype))
xw = x$Wortd - y$Wortd
xv = x$Vokd - y$Vokd
xc = x$Clustd - y$Clustd
x = data.frame(wd = xw, vd = xv, cld = xc, Vpn = x$Vpn, W = y$Wtype)
# so the durations in x should be bigger
# is duration of V in klick/knick > klicks, knicks?

par(mfrow=c(1,2))
# is duration of Cluster in klick/knick > klicks, knicks?
boxplot(cld * 1000 ~ W, data = x, main = "Clusterdauer", ylab = "Dauer (ms)")
abline(h=0, lty=2, col="blue")

boxplot(vd * 1000 ~ W, data = x, main = "Vokaldauer")
abline(h=0, lty=2, col="blue")

# boxplot((cld + vd) * 1000 ~ W, data = x, main = "Cluster+Vokaldauer")
# abline(h=0, lty=2, col="blue")

# compute difference values
temp = cb$Stype == "s1"
X = cb[temp,]
temp = cb$Stype == "s4"
Y = cb[temp,]
temp = cb$Stype == "s5"
Z = cb[temp,]

yw = X$Wortd - Y$Wortd
yv = X$Vokd - Y$Vokd
yc = X$Clustd - Y$Clustd
y = data.frame(wd = yw, vd = yv, cld = yc, Vpn = X$Vpn, W = Y$Wtype)


# is duration of Cluster in kl/nickt > kl/nickten?
boxplot(cld*1000 ~ W, data = y, main = "Clusterdauer", ylab= "Dauer (ms)")
abline(h=0, lty=2, col="blue")

# Is the duration of V bigger in kl/nickt than in kl/nickten
boxplot(vd*1000 ~ W, data = y, main = "Vokaldauer", ylab= "Dauer (ms)")
abline(h=0, lty=2, col="blue")




zw = X$Wortd - Z$Wortd
zv = X$Vokd - Z$Vokd
zc = X$Clustd - Z$Clustd
z = data.frame(wd = zw, vd = zv, cld = zc, Vpn = X$Vpn, W = Z$Wtype)


# is duration of Cluster in kl/nickt > gekl/nickten?
boxplot(cld*1000 ~ W, data = z, main = "Clusterdauer", ylab= "Dauer (ms)")
abline(h=0, lty=2, col="blue")

# Is the duration of V bigger in gekl/nickt than in gekl/nickten
boxplot(vd*1000 ~ W, data = z, main = "Vokaldauer", ylab= "Dauer (ms)")
abline(h=0, lty=2, col="blue")

# Wortdauer in kl/nickten, gekl/nickten
ylim = c(-100, 100)

boxplot(wd*1000 ~ W, data = y, main = "Wortdauer", ylab= "Dauer (ms)", ylim=ylim)
abline(h=0, lty=2, col="blue")

boxplot(wd*1000 ~ W, data = z, main = "Wortdauer", ylab= "Dauer (ms)", ylim=ylim)
abline(h=0, lty=2, col="blue")