# R Exercises

# 1  U.S. Education Data (simple and multiple regression)

library(car)
data(States)
attach(States)

summary(lm(SATM ~ percent))
summary(lm(SATM ~ dollars))
summary(mod.ed <- lm(SATM ~ percent + dollars))


# 2  Burt's IQ Data  (dummy regression)

detach(States)
data(Burt)
attach(Burt)
class <- factor(class, levels=c("low", "medium", "high"))

scatterplot(IQbio ~ IQfoster | class, smooth=FALSE)

mod.burt <- lm(IQbio ~ IQfoster*class)
summary(mod.burt)
Anova(mod.burt)

# 3  Powers and Xie's NLSY Data  (logistic regression)

detach(Burt)
Powers <- read.table(
    "http://socserv.socsci.mcmaster.ca/jfox/Courses/Oxford/Powers.txt",
    header=TRUE)
some(Powers)
dim(Powers)

mod.powers.1 <- glm(hsgrad ~ ., family=binomial, data=Powers)
summary(mod.powers.1)

mod.powers.2 <- update(mod.powers.1, . ~ . - nsibs + factor(nsibs))
summary(mod.powers.2)
anova(mod.powers.1, mod.powers.2, test="Chisq")

table(Powers$nsibs)

Powers$nsibs[Powers$nsibs < 0] <- NA
table(Powers$nsibs)

mod.powers.1.corrected <- glm(hsgrad ~ ., family=binomial, data=Powers)
mod.powers.2.corrected <- update(mod.powers.1.corrected, 
    . ~ . - nsibs + factor(nsibs))
anova(mod.powers.1.corrected, mod.powers.2.corrected, test="Chisq")


# 4  Long's Productivity Data  (Poisson regression)

Long <- read.table(
    "http://socserv.socsci.mcmaster.ca/jfox/Courses/Oxford/Long.txt",
    header=TRUE)
some(Long)
dim(Long) 

table(Long$art)
plot(sort(unique(Long$art)), table(Long$art), type="h")
points(sort(unique(Long$art)), table(Long$art), pch=16)

mod.long.1 <- glm(art ~ ., family=poisson, data=Long)
summary(mod.long.1)

mod.long.2 <- update(mod.long.1, family=quasipoisson)
summary(mod.long.2)