##############################
#
# Script: Solutions 10 (assignment10.R)
#
##############################


afg <- read.csv("http://www.electoralforensics.org/datasets/afg2009pres.csv", header=TRUE)
names(afg)

# Variables of interest:
#    Votes for:
#    . Dr. Abdullah Abdullah
#    . Hamed Karzai
#    Invalid Votes
#    Total Valid Votes
#    Total Votes
#    Province

# To get this information from a "long" dataset, you do the follwoing:
  v.abd <- afg$VOTES[afg$CANDIDATE=="Dr. Abdullah Abdullah"]
  v.kar <- afg$VOTES[afg$CANDIDATE=="Hamed Karzai"]
  v.inv <- afg$VOTES[afg$CANDIDATE=="Invalid Votes"]
  v.val <- afg$VOTES[afg$CANDIDATE=="Total Valid Votes"]
  v.tot <- afg$VOTES[afg$CANDIDATE=="Total Votes"]
  prov  <- afg$PROVINCE[afg$CANDIDATE=="Total Valid Votes"]


# Now, two variables of interest
  p.kar <- v.kar/v.tot
  w.kar <- as.numeric(v.abd<v.kar)

# Create this if you are using a Binomial
  y.inv <- cbind(v.inv,v.val)

# Initial model:
  model.1 <- glm(y.inv~p.kar*w.kar, family=binomial(link=logit) )
  summary(model.1)


# continue ...