##################
# dummy variable #
##################
# part 1
rm(list = ls()) 
ad = "https://www.fsb.miamioh.edu/lij14/411_wage.txt"
data = read.table(url(ad), header=T) 
attach(data)

prop.table(xtabs(~ female))
mean(female)

ybarm = mean(wage[female==0])
ybarf = mean(wage[female==1])
barplot(c(ybarm,ybarf), col=c("red", "blue"), names.arg = c("Male", "Female"))

# part 2
mean(wage[female==0])
mean(wage[female==1])
t.test(wage~female, var.equal = T)

summary(lm(wage~female))$coef

# part 3
plot(educ, wage, pch=16, col=ifelse(female==0,2,4))
abline(lm(wage[female==0]~educ[female==0]), col="red")
abline(lm(wage[female==1]~educ[female==1]), col="blue")

summary(lm(wage[female==0]~educ[female==0]))$coef
summary(lm(wage[female==1]~educ[female==1]))$coef

ss1 = subset(data, female==0)
summary(lm(wage~educ, data=ss1))$coef
ss2 = subset(data, female==1)
summary(lm(wage~educ, data=ss2))$coef

inta = educ*female
summary(lm(wage~female+educ+inta))$coef

# part 4
inta = exper*female
summary(lm(wage~female+exper+inta))$coef
m = lm(wage~female+exper+inta)
library("car")
linearHypothesis(m, c("female=0", "inta=0")) 

# part 5
sm = mean(wage[female==0&married==0])
sf = mean(wage[female==1&married==0])
mm = mean(wage[female==0&married==1])
mf = mean(wage[female==1&married==1])
DID = (mf-mm)-(sf-sm)
DID
inta = female*married
summary(lm(wage~female+married+inta))$coef

ca = rep(0,length(wage))
ca[female==0&married==0] = 1
ca[female==1&married==0] = 2
ca[female==0&married==1] = 3
ca[female==1&married==1] = 4
barplot(tapply(wage,ca,mean),names.arg = c("sm", "sf","mm", "mf")) 

############
# NBA data #
############
rm(list = ls()) 
ad = "https://www.fsb.miamioh.edu/lij14/311_nba.txt"
data = read.table(url(ad), sep="\t", header=T) 
attach(data)

position[1:5]
prop.table(xtabs(~position))

# Conditional Means
barplot(tapply(points,position,mean,na.rm=T))
mean(points[position=="center"],na.rm=T)
mean(points[position=="forward"],na.rm=T)
mean(points[position=="guard"],na.rm=T)

# Dummy Variables for each position
center = as.integer(position=="center")
forward = as.integer(position=="forward")
guard = as.integer(position=="guard")

# DVR and ANOVA (F test)
summary(lm(points~forward+guard))$coef
library("car")
m = lm(points~forward+guard)
linearHypothesis(m, c("forward=0", "guard=0")) 

# Dummy variable trap
summary(lm(points~center+forward+guard))$coef
summary(lm(points~center+forward+guard-1))$coef

# A shortcut
summary(lm(points~factor(position)))$coef


################
# mini project #
################
library(HistData)
data(GaltonFamilies)
attach(GaltonFamilies)

plot(father, childHeight, col =ifelse(gender=="male",1,2))
abline(lm(childHeight[gender=="male"]~father[gender=="male"]), col = 1)
abline(lm(childHeight[gender=="female"]~father[gender=="female"]), col = 2)

summary(lm(childHeight~father+gender))$coefficient

summary(lm(childHeight[gender=="male"]~father[gender=="male"]))$coefficient
summary(lm(childHeight[gender=="male"]~father[gender=="male"]))$r.squared

summary(lm(childHeight[gender=="female"]~father[gender=="female"]))$coefficient
summary(lm(childHeight[gender=="female"]~father[gender=="female"]))$r.squared