library(lattice)

grades <- read.table(file = "C:/Documents and Settings/steeleb/My Documents/Stat 341/NonameGrades.txt",header=T)
colnames(grades) <- c("H1","H2","H3","E1")
head(grades)
hmk <- rowSums(grades[,1:3])
exam <- grades$E1
mean(exam)
median(exam)
hist(exam,breaks=seq(from=0,to=101,by=5))

xyplot(exam~hmk,type=c("p","r"),col=1,pch=16)
I <- exam > 0

xyplot(exam[I]~hmk[I],type=c("p","r"),col=1,pch=16,xmin=0)
c(cor(exam,hmk),cor(exam[I],hmk[I]))
summary(lm(exam[I]~hmk[I]))

############## center x and y
x <- hmk[I] - mean(hmk[I])
y <- exam[I] - mean(exam[I])
xyplot(y~x,type=c("p","r"),col=1,pch=16,xmin=0)

summary(lm(y~x))
summary(lm(y~-1+x)) # eliminate intercept from model


####################################### melanoma

head(melanoma)
Year <- melanoma$year
Incidence <- melanoma$incidence
xyplot(Incidence~Year,type=c("p","r"),col=1,pch=16)
cor(Incidence,Year)

summary(lm(Incidence~Year))
Year.post36 <- Year - 1936
xyplot(Incidence~Year.post36,type=c("p","r"),col=1,pch=16)
cor(Incidence,Year.post36)
summary(lm(Incidence~Year.post36))


library(DAAG)
help("ais")
head(ais)

xyplot(ais$ssf~ais$pcBfat|ais$sex,col=1,xlab="Percent body fat",ylab="Skinfold thickness",pch=16)
cor(ais$ssf,ais$pcBfat)
I <- ais$sex=="f"
sum(I)
cor(ais$ssf[I],ais$pcBfat[I])


I <- ais$sex=="m"
sum(I)
cor(ais$ssf[I],ais$pcBfat[I])

######################### 
# 
# Investigate the extent to which hemoglobin concentration, or some other measure of 
# blood oxygen-carrying capacity (a measure of aerobic fitness) is associated with percent body fat
# (or some other measure of fat-to-muscle proportion). Be sure to examine sports individually.
# Which variables show the greatest degree of linear association? Which sports show the
# greatest degree of linear association? Can the association be explained if we consider the role of
# body weight towards performance of specific sports?
#
# To answer these questions, 
# (i) provide a multipanel graph using the variables you selected as most informative and hence 
#       showing a maximal degree of association
# (ii) report correlation coefficients for each sport
# (iii) Be sure to identify any sport with too few observation pairs to be informative in this analysis.
# 
###############################################

# Example

y <- ais$hg # choose a measure of aerobic fitness
x <- ais$lbm  # choose a measure of fat-to-muscle proportion
xyplot(y~x,groups=ais$sex,xlab="x",ylab="y",type=c("p","r"),pch=16,col=c("red","blue"))
legend("topright",col=c("red","blue"),pch=16,legend=levels(ais$sex))

xyplot(y~x|ais$sport,col=1,xlab="x",ylab="y",type=c("p","r"),pch=16)
xyplot(y~x|ais$sport,groups=ais$sex,xlab="x",ylab="y",type=c("p","r"),pch=16,col=c("red","blue"))
legend("topright",col=c("red","blue"),pch=16,legend=levels(ais$sex))

v <- length(levels(ais$sport)) # number of levels of sport
for (i in 1:v){
	sport.i <- levels(ais$sport)[i] # select a sport
	I <- ais$sport==sport.i
	print(sport.i)
	print(c(sum(I),cor(x[I],y[I])))
}