Basic statistics on dutch election 2010

dutch.elections.2010.R

# read in ducth election file as provided by http://www.engagedata.eu/dataset/14399
# https://engagefp7.s3.amazonaws.com/resources/dataset_14399/TK2010.csv

dat = read.csv2('http://engagefp7.s3.amazonaws.com/resources/dataset_14399/TK2010.csv', skip=31, header=FALSE)

#define column names
columns =  '
Gemeente; 
geldige stemmen; 
ongeldige stemmen; 
blanco stemmen;
Percentage blanco stemmen; 
kiesgerechtigden;
Opkomst;
Opkomstpercentage;
VVD;
Partij van de Arbeid (P.v.d.A.);
PVV (Partij voor de Vrijheid);
Christen Democratisch Appel (CDA);
SP (Socialistische Partij);
Democraten 66 (D66);
GROENLINKS;
ChristenUnie;
Staatkundig Gereformeerde Partij (SGP);
Partij voor de Dieren;
TROTS OP NEDERLAND LIJST RITA VERDONK;
Partij voor Mens en Spirit (MenS);
Piratenpartij;
Lijst 17;
Partij een;
Nieuw Nederland;
Heel NL;
Evangelische Partij Nederland
'
#parse the names as a dataframe
column.names = read.csv2(text=columns, header=FALSE)

#rename the columns of the data frame
colnames(dat) <- column.names[,1]

#remove last four lines
dat = dat[1:(nrow(dat)-4),]

#how many municipalities?
nrow(dat)
length(dat$Gemeente)
length(unique(dat$Gemeente))

#how many parties
n    = names(dat)
from = which(names(dat) == 'Opkomstpercentage') + 1
to   = length(n)

#select the columns relative to parties
parties = n[from:to]
parties

#how many parties?
length(parties)

#votes per party
total = colSums(dat[parties])
total

#plot bars
barplot(total)

#plot log10 (some alternatives)
barplot(log10(total))
barplot(total, log='y')

# most votes
total.max.value = max(total)
total.max       = total[total==total.max.value]
total.max

#minum votes
total.min.value = min(total)
total.min       = total[total==total.min.value]
total.min

# coverage min votes per municipality

#at least one vote for the looser
I = dat[names(total.min)] > 0

#exactly 12 votes for the looser
J = dat[names(total.min)] == 12

#select the municipalities for the loosing party (at least one vote)
party.loosing = dat[I,c('Gemeente', names(total.min))]
nrow(party.loosing)

#select the municipalities for the loosing party (exactly 12 votes)
party.loosing = dat[J,c('Gemeente', names(total.min))]
nrow(party.loosing)

# CDA < VVD < PVV

I = dat['VVD'] < dat['PVV (Partij voor de Vrijheid)']
J = dat['VVD'] > dat['Christen Democratisch Appel (CDA)']

d = dat[I & J, c('Gemeente', 'VVD', 'PVV (Partij voor de Vrijheid)', 'Christen Democratisch Appel (CDA)')]

#which municipalities where  CDA < VVD < PVV
d$Gemeente
nrow(d)

# time for pies!
pie(total)

# create a others category for the small parties
n = 7
sorted.total = sort(total, decreasing=TRUE)

votes = sorted.total[1:n]
votes = c(votes, others=sum(sorted.total[(n+1):length(total)]))

# a better pie
pie(votes, main='dutch elections 2010')

# as above but for a specific municipality
municipality = 'Almere'

municipality.selection = dat$Gemeente == municipality
municipalities.votes = dat[municipality.selection, parties]

# reuse previous code ...
total = as.numeric(municipalities.votes)
names(total) = parties

n = 7
sorted.total = sort(total, decreasing=TRUE)

votes = sorted.total[1:n]
votes = c(votes, others=sum(sorted.total[(n+1):length(total)]))

# a better pie
pie(votes, main=municipality)

# take into account valid votes and correct the votes by valid votes
relative = dat[parties] / dat[,'geldige stemmen']

# relative correlation
cross.corr = cor(relative)

# do not look in the diagonal
diag(cross.corr) = NA

# VVD correlation
vvd.corr = cross.corr['VVD',]
vvd.corr

# min and max correlation for vvd
vvd.corr.max = max(vvd.corr, na.rm=TRUE)
vvd.corr.max

vvd.corr.min = min(vvd.corr, na.rm=TRUE)
vvd.corr.min

# which party correlates?
# positive correlation
vvd.corr[which.max(vvd.corr)]

# which party correlates?
# negative correlation
vvd.corr[which.min(vvd.corr)]

#get the names of the parties (row and column are the same)
cross.corr.parties = rownames(cross.corr)

# max correlation (all columns)
corr.max.value = apply(cross.corr, 2, FUN=function(x) {max(x, na.rm=TRUE)} )
corr.max.value

corr.max.party = apply(cross.corr, 2, FUN=function(x) {cross.corr.parties[which.max(x)]} )
corr.max.party

#combine into a data frame
cross.corr.dataframe = data.frame( cbind(corr.max.value, corr.max.party) )
cross.corr.dataframe

#sort by highest correlation
I = order(cross.corr.dataframe[,1],decreasing=TRUE)
cross.corr.dataframe[I,]

#highest correlation
cross.corr.dataframe[I[1],]


# min correlation (all columns)
corr.min.value = apply(cross.corr, 2, FUN=function(x) {min(x, na.rm=TRUE)} )
corr.min.value

corr.min.party = apply(cross.corr, 2, FUN=function(x) {cross.corr.parties[which.min(x)]} )
corr.min.party

#combine into a data frame
cross.corr.dataframe = data.frame( cbind(corr.min.value, corr.min.party) )
cross.corr.dataframe

#sort by highest negative correlation
I = order(cross.corr.dataframe[,1],decreasing=TRUE)
cross.corr.dataframe[I,]

#highest correlation
cross.corr.dataframe[I[1],]