library(sasld)

# 3.1 
data(pesticidi)

pesticidi$status <- as.character(pesticidi$status)
table(pesticidi$food,pesticidi$status)

Table(pesticidi$food,pesticidi$status)

tbl <- table(pesticidi$food,pesticidi$status)
prop.table(tbl,margin=1)

# "accorcio" il nome delle modalità
status <- pesticidi$status
status[pesticidi$status == "Presente"] <- "Pres."
status[pesticidi$status == "Assente"]  <- "Ass."
pesticidi$status <- status

tc(pesticidi$food,pesticidi$status)
# -------------------------------------------------------------------




# 3.2 
data(www)

internet <- www[,10]
facebook <- www[,11]
cor(internet,facebook)




# parabola
a <- -3; b <- 24; c <- 80
x <- seq(2,6,0.5)
y <- a*x^2 + b*x + c
cbind(x,y)

cor(x,y)

plot(x,y)
-b/(2*a)

plot(x[-c(1,2)],y[-c(1,2)],xlab="x",ylab="y")
cor(x[-c(1,2)],y[-c(1,2)])

cor(x[-c(8,9)],y[-c(8,9)])


# non lineare
x <- c(0,1,2,3,4,5,6)
y <- c(1,2,4,8,16,32,64)
cor(x,y)

cor(x,log(y))

cor(x,log10(y))




# outlier
set.seed(654123)
x <- rnorm(50,170,10)
y <- rnorm(50,170,10)

cor(x,y)
plot(x,y,pch=19)

xx <- c(x,300); yy <- c(y,300)
cor(xx,yy)
plot(xx,yy,pch=19)

xx <- c(x,300); yy <- c(y,10)
cor(xx,yy)
plot(xx,yy)


#
data(ballot)

plot(ballot$perot,ballot$buchanan)
cor(ballot$perot,ballot$buchanan)
cor(ballot$perot[-50],ballot$buchanan[-50])

out <- numeric(nrow(ballot))
for (i in 1:nrow(ballot)) out[i] <- cor(ballot$perot[-i],ballot$buchanan[-i])
round(out,4)

cor(ballot[,-1])
# -------------------------------------------------------------------




# 3.3 
data(crime)

plot(crime$college,crime$murder.rate)

fit <- lm(murder.rate ~ college, data=crime)
fit

plot(crime$college,crime$murder.rate)
abline(fit)

fit.1 <- lm(murder.rate ~ college, data=crime)
fit.2 <- lm(murder.rate[-9] ~ college[-9], data=crime)
plot(crime$college,crime$murder.rate)
abline(fit.1,lty=2,lwd=2)
abline(fit.2,lty=3,lwd=2)
# -------------------------------------------------------------------




# 3.4 
data("flCrime")

cor(flCrime$education,flCrime$crime)

cor(data.frame(educ=flCrime$education,crime=flCrime$crime,urb=flCrime$ urbanization))

fit.1 <- lm(crime ~ urbanization, data=flCrime)
fit.2 <- lm(education ~ urbanization, data=flCrime)
cor(fit.1$residuals,fit.2$residuals)
# -------------------------------------------------------------------