####################################################################################
#This is a file with executable R code for chapter 14 of Natalia Levshina's (2015) 
#How to Do Linguistics with R. Amsterdam/Philadelphia: John Benjamins. 
####################################################################################

###Section 14.2


##Main text

install.packages("party")

library(Rling); library(party)

data(caus)

mhc <- caus[caus$Cx == "make_V"|caus$Cx == "have_V"|caus$Cx == "cause_toV", ]

summary(mhc)

mhc$Cx <- factor(mhc$Cx)

summary(mhc$Cx)

levels(mhc$Cx) <- c("cause", "have", "make")

summary(mhc$Cx)

set.seed(129)

mhc.ctree <- ctree(Cx ~ CrSem + CeSem + CdEv + Neg + Coref + Poss, data = mhc)

plot(mhc.ctree)

table(predict(mhc.ctree), mhc$Cx)

(35 + 42 + 24)/150

set.seed(35)

mhc.rf <- cforest(Cx ~ CrSem + CeSem + CdEv + Neg + Coref + Poss, data = mhc, controls = cforest_unbiased(ntree = 1000, mtry = 2))

mhc.varimp <- varimp(mhc.rf, conditional = TRUE)

round(mhc.varimp, 3)

dotchart(sort(mhc.varimp), main = "Conditional importance of variables")

table(predict(mhc.rf), mhc$Cx)

(39 + 48 + 13)/150


##Boxes with additional information

mhc.ctree1 <- ctree(Cx ~ CrSem + CeSem + CdEv + Neg + Coref + Poss, data = mhc, controls = ctree_control(testtype = "MonteCarlo", mincriterion = 0.99, minbucket = 15))