##############################
#
#
# December 6, 2011
# Topics:
#   Importing data from other formats
#   Mid-Semester Review III
#
#
##############################


# Preamble

# We will need these packages today
library(RFS)
library(foreign)


# Small practice with the foreign package

# There are several file formats R can access, beyond the 
# non-proprietary, standard csv format. 

# Stata file format (.dta)
data1 <- read.dta("http://courses.kvasaheim.com/pols6123/data/test.dta")
summary(data1)

# SPSS (PSAW) file format (.sav)
data2 <- read.spss("http://courses.kvasaheim.com/pols6123/data/test.sav")
summary(data2)



# SAS is a more complex file system, but it can also be read by R:
#    read.xport 
#    read.ssd

# Weka file
#    read.arff

# Several different database programs like dBase and FoxPro
#    read.dbf

# etc ...


# The documentation for the package is located at:
#    http://cran.r-project.org/web/packages/foreign/foreign.pdf









rm(list=ls())

##################################################
# Activity 3.1: Patrick Henry


# Preamble
library(RFS)


# Read data
ph <- read.csv("http://courses.kvasaheim.com/pols6123/data/patrickHenry.csv")
attach(ph)


# Get to know the data
names(ph)
summary(ph)
head(ph)
str(ph)


# Preliminary modeling ... what works?
m1 <- glm( logit(gpa/4) ~ composite+highschool+gender, family=gaussian(link=identity))
summary(m1)

m10 <- glm( gpa/4 ~ composite+highschool+gender, family=Gamma(link=inverse))
summary(m10)

highschool <- set.base(highschool, "Public High School")
m10 <- glm( gpa ~ composite+highschool+gender, family=Gamma)
summary(m10)


m3 <- glm( gpa/4 ~ composite+highschool+gender, family=gaussian(link=make.link("logit")) )
summary(m3)

m4 <- glm( gpa/4 ~ composite*highschool*gender, family=gaussian(link=make.link("logit")) )
summary(m4)




# Predictions for two students

Bud   <- data.frame(highschool="Public High School", gender="Male",   composite=1600)
Kelly <- data.frame(highschool="Public High School", gender="Female", composite= 400)

logistic( predict(m3, newdata=Bud  ) ) *4   # Bud's predicted GPA
logistic( predict(m3, newdata=Kelly) ) *4   # Kelly's predicted GPA



# Prediction graph

newComposite <- seq(400,1600, length=1000)
prMV <- 4* logistic( predict(m3, newdata=data.frame(composite=newComposite, gender="Male",   highschool="Private High School") ) )
prFV <- 4* logistic( predict(m3, newdata=data.frame(composite=newComposite, gender="Female", highschool="Private High School") ) )
prMU <- 4* logistic( predict(m3, newdata=data.frame(composite=newComposite, gender="Male",   highschool="Public High School") )  )
prFU <- 4* logistic( predict(m3, newdata=data.frame(composite=newComposite, gender="Female", highschool="Public High School") )  )
prMH <- 4* logistic( predict(m3, newdata=data.frame(composite=newComposite, gender="Male",   highschool="Home School") )         )
prFH <- 4* logistic( predict(m3, newdata=data.frame(composite=newComposite, gender="Female", highschool="Home School") )         )

par(mar=c(4,4,0,0)+0.35)
plot(composite, gpa,  las=1, ylim=c(0,4), xlim=c(300,1700), xlab="SAT Composite Score", ylab="Student GPA")
lines(newComposite,prMV, lwd=2, col=4, lty=1)
lines(newComposite,prMU, lwd=2, col=4, lty=2)
lines(newComposite,prMH, lwd=2, col=4, lty=3)
lines(newComposite,prFV, lwd=2, col=2, lty=1)
lines(newComposite,prFU, lwd=2, col=2, lty=2)
lines(newComposite,prFH, lwd=2, col=2, lty=3)

legend("topleft", legend=c("Male, Public", "Male, Private", "Male, Home", "Female, Public", "Female, Private", "Female, Home"),
	lwd=2, col=c(4,4,4,2,2,2), lty=1:3, bty="n")