#vcd Paket
install.packages(c("vcd"),"/Library/Frameworks/R.framework/Resources/library",type="mac.binary");.refreshHelpFiles()
install.packages(c("colorspace"),"/Library/Frameworks/R.framework/Resources/library",type="mac.binary");.refreshHelpFiles()
library(vcd)

#Punishment Datensatz in R
data(Punishment)
names(Punishment)
attach(Punishment)
Punishment

#Tabellen
table1<-xtabs(Freq ~ memory + attitude + age + education)
table2<-xtabs(Freq~age+attitude)

#Lineare Modelle
length(attitude)
att<-rep(0,36)
att[attitude=="moderate"]<-1
mlm1<-lm(att~memory+education+age,weights=Freq)
par(mfrow=c(2,2))
plot(mlm1)

#Datensatz fŸr logistische Regression vorbereiten
Dp<-Punishment
Dp1<-Dp[1:18,3:5]
Dp0<-cbind(Dp[19:36,1],Dp[1:18,1])
DpX<-cbind(Dp0,Dp1)

attach(DpX)
names(DpX)[1]<-"Moderate"
names(DpX)[2]<-"No"

#logistische Regression Modelle
m1<-glm(Dp0~memory+education+age,family=binomial)
summary(m1)
m2<-glm(Dp0~memory+education,family=binomial)
summary(m2)
m3<-glm(Dp0~memory*education,family=binomial)
summary(m3)
anova(m3)

#Probit Modelle
m2a<-glm(Dp0~memory+education,family=binomial(link=probit))
summary(m2a)

#Vergleich
f2<-m2$fitted.values
f2a<-m2a$fitted.values
plot(f2,f2a)
dd<-f2-f2a
dd

plot(m2$y,f1)

#Vergleich von Linkfunktionen
bb<-seq(-3,3,0.001)
lr<-1/(1+exp(-bb))
pr<-pnorm(bb)
plot(bb,lr)
lines(bb,pr,type="l")
plot(bb,pr-lr,type="l")

#Besser die inversen Funktionen zu vergleichen?
cc<-seq(0.01,0.99,0.01)
lq<-log(cc/(1-cc))
pq<-qnorm(cc)
plot(cc,lq)
lines(cc,pq,type="l")

detach(DpX)

#loglineare Modelle
attach(Dp)
m4<-glm(Freq~attitude+memory*education*age,family=poisson)
m5<-glm(Freq~attitude*memory*education+age,family=poisson)
m6<-glm(Freq~attitude*memory+attitude*education+age,family=poisson)
m7<-glm(Freq~attitude*memory+attitude*education+attitude*age,family=poisson)
m8<-glm(Freq~attitude*memory+attitude*education+memory*education*age,family=poisson)