#
# This software has been released to the Public Domain,
# for a license text see 
# http://creativecommons.org/licenses/publicdomain/
#
# You need GNU R to use this software, see
# http://www.r-project.org/
#


# The data file ooxmlvote.csv contains the following columns in comma
# separated format:
# "Country"   Name of the country
# "CPI06"     Transparency International's Corruption Perceptions Index
#             (CPI) 2006, see
#             http://www.transparency.org/policy_research/surveys_indices/cpi/2006
# "CPI07"     Transparency International's Corruption Perceptions Index
#             (CPI) 2007, see
#             http://www.transparency.org/policy_research/surveys_indices/cpi/2007
# "Member"    Member organization in a counntry
# "Status"    Status of the member: P member (incl. Secretariat) or O member
# "Vote07"    Vote in 2007 on acceptance of OOXML
# "Vote08"    Vote in 2008 on acceptance of OOXML
# "Comment07" Comment file was submitted with the 2007 vote (TRUE/FALSE)
# "Comment08" Comment file was submitted with the 2008 vote (TRUE/FALSE)

V <- read.csv("http://www.effi.org/system/files?file=ooxmlvote.csv.txt")

# P-Members include the Secretariat country (United States).
pmembers <- (!is.na(V[,"Status"]) &
             (V[,"Status"]=="Secretariat" |
              V[,"Status"]=="P Member"))


##################################################################
# Vote on 2 September 2007
#
# P-Members voting 17 in favour out of 32 = 53% (requirement 66.66%)
table(V[pmembers,"Vote07"])
# Abstention    Approval Disapproval 
#          9          17          15 
# Note 17+15=32 (abstentions are not counted)

# Wilcoxon rank sum test for the hypothesis that countries that
# were perceived corrupt were more likely to vote for approval
# than disapproval. Statistical significance found (alpha=0.05).
wilcox.test(V[pmembers & V[,"Vote07"]=="Approval","CPI07"],
            V[pmembers & V[,"Vote07"]=="Disapproval","CPI07"],
            alternative="less")
# W = 77, p-value = 0.02921
# alternative hypothesis: true location shift is less than 0 



# Member bodies voting: 18 negative votes out of 69 = 26% (requirement 25%)
table(V[,"Vote07"])
# Abstention    Approval Disapproval 
#         18          51          18 
# Note: 51+18=69 (abstentions are not counted)

# Wilcoxon rank sum test for the hypothesis that countries that
# were perceived corrupt were more likely to vote for approval
# than disapproval. Statistical significance found (alpha=0.05).
wilcox.test(V[V[,"Vote07"]=="Approval","CPI07"],
            V[V[,"Vote07"]=="Disapproval","CPI07"],
            alternative="less")
# W = 277.5, p-value = 0.008354
# alternative hypothesis: true location shift is less than 0 


##################################################################
# Vote on 2 April 2008
#
# P-Members voting 24 in favour out of 32 = 75% (requirement 66.66%)
table(V[pmembers,"Vote08"])
# Abstention    Approval Disapproval 
#          9          24           8 
# Note 24+8=32 (abstentions are not counted)

# Wilcoxon rank sum test for the hypothesis that countries that
# were perceived corrupt were more likely to vote for approval
# than disapproval. No statistical significance.
wilcox.test(V[pmembers & V[,"Vote08"]=="Approval","CPI07"],
            V[pmembers & V[,"Vote08"]=="Disapproval","CPI07"],
            alternative="less")
# W = 119.5, p-value = 0.8523
# alternative hypothesis: true location shift is less than 0 


# Member bodies voting: 10 negative votes out of 71 = 14% (requirement 25%)
table(V[,"Vote08"])
# Abstention    Approval Disapproval 
#         16          61          10 
# Note: 61+10=71 (abstentions are not counted)

# Wilcoxon rank sum test for the hypothesis that countries that
# were perceived corrupt were more likely to vote for approval
# than disapproval. No statistical significance.
wilcox.test(V[V[,"Vote08"]=="Approval","CPI07"],
            V[V[,"Vote08"]=="Disapproval","CPI07"],
            alternative="less")
# W = 325, p-value = 0.6658
# alternative hypothesis: true location shift is less than 0 


##################################################################
# Comparing 2007 and 2008 votes

# The P-Member vote:
table(V[pmembers,"Vote07"],V[pmembers,"Vote08"])
#              Abstention Approval Disapproval
#  Abstention           6        3           0
#  Approval             2       14           1
#  Disapproval          1        7           7

# All votes:
table(V[,"Vote07"],V[,"Vote08"])
#              Abstention Approval Disapproval
#  Abstention          11        7           0
#  Approval             4       45           2
#  Disapproval          1        9           8

##################################################################
# Figure

x <- matrix(c(table(factor(floor(V[V[,"Vote08"]=="Approval","CPI07"]),
                           levels=2:9)),
               table(factor(floor(V[V[,"Vote08"]=="Disapproval","CPI07"]),
                            levels=2:9))),
            nrow=2,ncol=8,byrow=TRUE,
            dimnames=list(c("approval","disapproval"),2:9))
barplot(x,
        main='"10 negative votes out of 71"',
        xlab="lots of corruption « CPI » little corruption",
        ylab="national body votes (except Fiji)",
        legend.text=c("approval","disapproval"))