#Creating the dichotomous variable for high income.
depress$inc.high <- 1*(income > 19)

#Table 17.1, p. 412.
attach(depress)

depress$sex.fac <- factor(sex, labels=c("male", "female"))
depress$inc.high.fac <- factor(inc.high, labels=c("low", "high"))

table(depress$sex.fac, depress$inc.high.fac)

#Creating the factor variable for treat.
depress$treat.fac <- factor(treat, labels=c("No", "Yes"))
depress$cases.fac <- factor(cases, labels=c("low", "high"))

#Table 17.2, p. 413,
table(depress$sex.fac, depress$inc.high.fac, depress$treat.fac)

#Creating the factor variable for cases.
depress$cases.fac <- factor(cases, labels=c("low", "high"))

#Table 17.3, p. 414.
table(depress$sex.fac, depress$inc.high.fac, depress$treat.fac, 
      depress$cases.fac)

#Output for chi-squared test, bottom p. 419.
chisq.test(depress$inc.high, depress$sex, correct=F)

#Table 17.7, p. 420.
table1 <- table(depress$sex, depress$inc.high)
loglin(table1, list(1:2), param=T, fit=T)

#output for three-way model, p. 435-437.
table3 <- table(depress$sex, depress$inc.high, depress$treat)
loglin(table3, list(1:3), param=T, fit=T)

#Calculations, top p. 437.
#Note we need to be modeling a (0, 1) variable rather than a (1, 2)
#variable.  Also, we want the predictor sex to be (0, 1) rather (1, 2).
depress$sex1 <- depress$sex - 1
depress$treat1 <- depress$treat -1

glm(treat1 ~ sex1 + inc.high, depress, family=binomial)

#Detaching the data set if done.
detach(depress)