#
#  MATH3330 Section B    October 2001
#  Week5b_Coffee_Part1
#
#

coffee <- read.table('coffee.dat', header = T, row.names = NULL, skip = 2)

# Beware: this is 'artificial' data. Don't draw any conclusions from it

summary(coffee)
plot( coffee , ask = F)

pairs( coffee )

xyplot( Heart ~ Coffee , coffee,
	panel = function(x,y,...) {
		panel.xyplot(x,y,...)
		panel.lmline(x,y,...)
	})
	

occupation <- c(rep('Student',8), rep('Professor',12))


identify( 
	xyplot( Heart ~ Coffee , coffee,
		panel = function(x,y,...) {
		panel.xyplot(x,y,...)
		panel.lmline(x,y,...)
	}), 
	labels = occupation
)

fit.c <- lm( Heart ~ Coffee, coffee)
summary(fit.c)  # very strong positive relationship between Heart Damage and Coffee drinking


# How about stress

fit.s <- lm( Heart ~ Stress, coffee)
summary(fit.s)  # very strong positive relationship between Heart Damage and Stress


# So if we use both we would expect to see a strong relationship with BOTH!

fit.cs <- lm( Heart ~ Coffee + Stress, coffee)
summary(fit.cs) 

# BUT the relationship with coffee is negative!!

summary( coffee )

pred <- expand.grid ( Coffee = seq( 25, 150, by = 5), Stress = seq( 15, 200, by=5 ))

pred.c <- pred
pred.c$Heart <- predict(fit.c, newdata = pred)

pred.s <- pred
pred.s$Heart <- predict(fit.s, newdata = pred)

pred.cs <- pred
pred.cs$Heart <- predict(fit.cs, newdata = pred)

pred <- stack( coffee, pred.c, pred.s, pred.cs, Type. = c('Data','Coffee only','Stress only','Stress+Coffee'))
cloud( Heart ~ Coffee + Stress | Type. , pred)