% \VignetteIndexEntry{Duration of Unemployment - Trees}
% \VignetteDepends{party}
%\VignetteEngine{knitr::knitr}
%\VignetteEncoding{UTF-8}

\documentclass[a4paper]{article}

\title{Duration of Unemployment - Trees}

\begin{document}

\maketitle


<<echo=FALSE,eval=FALSE>>=
options(width=80)
@

<<eval=FALSE>>=
library(catdata)
data(unemployment,package="catdata")
@

To fit a tree for the unemployment data we use "ctree" from the library "party". 

<<fig.width=9,eval=FALSE>>=
library(party)
tree1<-ctree(as.factor(durbin)~age,data=unemployment)

plot(tree1)
@

The fitted regression function can be obtained by computing the respective means within the identified regions and plot them as function of age.

<<eval=FALSE>>=
unemployment$durbin[unemployment$durbin==2]<-0
year<- unemployment$age
year [unemployment$age<29.5] <- 1
year [unemployment$age>29.5 & unemployment$age<52.5] <- 2
year [unemployment$age>52.5] <- 3
 	
pre3 <- mean(unemployment$durbin[year==3])
pre2 <- mean(unemployment$durbin[year==2])
pre1 <- mean(unemployment$durbin[year==1])

meanyear <- c()

for (i in min(unemployment$age):max(unemployment$age)){
meanyear[i] <- sum(unemployment$durbin[unemployment$age==i])
if(sum(unemployment$durbin[unemployment$age==i])!=0){
meanyear[i] <- mean(unemployment$durbin[unemployment$age==i])
}
}

unemployment$means<- rep(2, nrow(unemployment)) 
 
 for (k in 1:nrow(unemployment)){
 unemployment$means[k] <- meanyear[unemployment$age[k]]
 }
@

<<fig.width=9,eval=FALSE>>=
  plot(unemployment$age, unemployment$means, xlab="age",ylab="",cex.axis=1.5,
       cex.lab=1.5)
 segments(x0=min(unemployment$age),x1=29.5,y0=pre1)
 segments(x0=29.5,x1=29.5,y0=pre1,y1=pre2)
 segments(x0=29.5,x1=52.5,y0=pre2)
 segments(x0=52.5,x1=52.5,y0=pre2,y1=pre3)
 segments(x0=52.5,x1=max(unemployment$age),y0=pre3)
@

<<echo=FALSE,results='hide',eval=FALSE>>=
detach(package:party)
@
\end{document}