Example Exploratory Data Analysis

This is an unmarked optional tutorial to show the kind of thinking that goes into an exploratory data analysis
The goal of this tutorial document is to walk through some of the common issues encountered in the early
stages of an exploratory analysis on a set of data. It gives examples of common problem areas in:
reading in data
dealing with blanks
dealing with factors
This data is a modied version of data from the New Zealand Election Survey, deliberately modied to
introduce problems that occur naturally in many data sets.

Step One. Learn something about the data set.

In this case, the New Zealand Election Survey takes place every three years as a postal survey of a sample of
registered electors. Some sampled electors were part of a sample panel of people surveyed at the previous
election as part of a longitudinal study, others were randomly chosen from the electoral roll. Those electors
that were part of the longitudinal panel group were randomly selected in previous elections.
As well as survey results, the data set includes information from the electoral roll, and weighting values for
adjusting results. The full NZES data set has been reduced to a selected group of variables, making 3101
observations of 107 variables.

Step Two. Contemplate some questions.

Examining the codebook (or in this case the appendix at the end of the document to check out the variables).
For example, we might decide that since New Zealand is a Mixed Member Proportional voting system, where
people get to vote for both an electorate (local) representative and a nationwide political party, that it would
be interesting to look at strategic voting under conditions where there are many political parties to choose
from. We identify some relevant variables of interest in the data, and investigate the nature of the individual
variables before we explore their interactions. The kind of variables they are is going to shape our question.

Read in the data

There are many dierent kinds of data les in the world. Each one has its own issues when being read in by
R. In this case the data is saved as a .RData le, which can be read in by using the load() command.
As with most reading in le commands, inside the parentheses needs to go a piece of text, in quotes, that is
the path to the le from the working directory (the working directory is the folder that R is currently paying
attention to). The easiest way to get a R Markdown (Rmd) Document and console cooperating about this is to
place the le with the data in it in the same folder as the R Markdown (Rmd) Document, open the R
Markdown (Rmd) document in RStudio so we are looking at the contents of the document in the editing
window, then in the RStudio Session menu, use the Set Working Directory - To Source File Location
command to make a common starting point. Then the code in the R Markdown (Rmd) document will use the
same working space regardless of whether we knit the document or run code chunks in the Console. In this
case, if the nzes2011.RData le is in the same folder as the R Markdown (Rmd) document, hence it can be
read into R with the following command:

load("selected_nzes2011.Rdata")

We also want to load packages that have functions in them we want to use. For this particular analysis we
will only need the dplyr package, but for your project you will also likely need other packages as well, e.g.
ggplot2 .
library(dplyr)

Step Three. Prepare for the rst question

As a rst question, we might be interested in exploring the relationship between the party the person voted
for, the party that was their favourite, and if they believed that their vote makes a dierence focusing on
the question that are people who believe their vote makes a dierence more likely to strategically vote for a
party not their favourite. To achieve this, we familiarise ourselves with the variables jpartyvote ,
jdiffvoting , and _singlefav . First we check the codebook (see Appendix), then we explore the data.
Viewing the entire dataset in the Data Viewer window by clicking on the data frames name in the
Environment or running the View() command in the Console can be ineective since the Data Viewer only
shows the rst 100 columns of the data frame.
Using the str() command on the entire dataset can also be equally ineective. However we can subset the
columns of interest and take a closer look at them. We can use the dplyr chain to select the variables of
interest and investigate only their structure by adding str() at the end of the chain:
selected_nzes2011%>%
select(jpartyvote,jdiffvoting,_singlefav)%>%
str()

If we try to run that line, we will get an error message about unexpected input or missing object.
We next need to diagnose where the problem lies in the R code or in the data? The best way to
troubleshoot this issue is to run each line of the dplyr chain one by one.
selected_nzes2011

The rst line runs without any erros, but the second line gives an error
selected_nzes2011%>%
select(jpartyvote,jdiffvoting,_singlefav)

We know that select() is a valid dplyr function, so that cannot be the problem. This means the problem
might be the variable names. The issue is that R has rules about what variable names are legal (e.g.no
spaces, starting with a letter) and when data is loaded, R will often x variable names to make them legal.
This happened to the _singlefav at the time of loading the data.
We could check this by looking through every single variable name in the data with the names() command.
names(selected_nzes2011)

##[1]"Jelect""jblogel""jnewspaper""jnatradio"
##[5]"jtalkback""jdiscussp""jrallies""jpersuade"
##[9]"jpcmoney""jpcposter""jlablike""jnatlike"
##[13]"jgrnlike""jnzflike""jactlike""junflike"
##[17]"jmaolike""jmnplike""jmostlike""jmostlikex"
##[21]"jrepublic""jsphealth""jspedu""jspunemp"
##[25]"jspdefence""jspsuper""jspbusind""jsppolice"
##[29]"jspwelfare""jspenviro""jgovpdk""jgovplab"
##[33]"jgovpnat""jgovpgrn""jgovpnzf""jgovpact"
##[37]"jgovunf""jgovpmao""jgovpmnp""jnevervoteno"
##[41]"jnevervotelab""jnevervotenat""jnevervotegrn""jnevervotenzf"
##[45]"jnevervoteact""jnevervoteunf""jnevervotemao""jnevervotemnp"
##[49]"jnevervoteoth""jnevervoteothx""jfirstpx""jsecondp"
##[53]"jage""jlanguage""jlanguagex""jrollsex"
##[57]"jhqual""jwkft""jwkpt""jwkun"
##[61]"jwkret""jwkdis""jwksch""jwkunpo"
##[65]"jwkunpi""jhhincome""jhhadults""jhhchn"
##[69]"jmarital""r_jind""jlablr""jnatlr"
##[73]"jgrnlr""jnzflr""jactlr""junflr"
##[77]"jmaolr""jmnplr""jslflr""jrelservices"
##[81]"jrelnone""jrelang""jrelpres""jrelcath"
##[85]"jrelmeth""jrelbap""jrellat""jrelrat"
##[89]"jrelfun""jrelothc""jrelnonc""jreligionx"
##[93]"jreligiousity""jethnicity_e""jethnicity_m""jethnicity_p"
##[97]"jethnicity_a""jethnicity_o""jethnicityx""jethnicmost"
##[101]"jethnicmostx""jpartyvote""jelecvote""njptyvote"
##[105]"njelecvote""jdiffvoting""X_singlefav"

However, when we have hundreds of column names, a useful tip is to just search out only possible names.
We can search the names for a fragment of the name by using the
grep("FRAGMENT",variable,value=TRUE) command, which in this case might be:
grep("singlefav",names(selected_nzes2011),value=TRUE)

##[1]"X_singlefav"

The value=TRUE argument, as described in the help for the grep() function reports the mathing character
string, as opposed to the index number for that string.
We can now conrm that the variable is called X_singlefav , so that is how we should be referring to it.
selected_nzes2011%>%
select(jpartyvote,jdiffvoting,X_singlefav)%>%
str()

These are all categorical data, however they are recorded as characters (text strings) as opposed to factors.
An easy way of tabulating these data to see how many times each level of is to use the group_by() function
along with the summarise() command:

selected_nzes2011%>%
group_by(jpartyvote)%>%
summarise(count=n())

###Atibble:142
##jpartyvotecount
##<chr><int>
##1Act29
##2ALC10
##3Alliance2
##4Anotherparty8
##5Conservative74
##6Don'tknow23
##7Green348
##8Labour749
##9Mana62
##10MaoriParty128
##11National1130
##12NZFirst216
##13UnitedFuture14
##14<NA>308

We can see that 23 people answered "Don'tknow" . Since our question is about people who knew which
party they voted for, we might want to exclude these observations from our analysis. We can do so by
filter ing them out.
selected_nzes2011%>%
filter(jpartyvote!="Don'tknow")%>%
group_by(jpartyvote)%>%
summarise(count=n())

###Atibble:122
##jpartyvotecount
##<chr><int>
##1Act29
##2ALC10
##3Alliance2
##4Anotherparty8
##5Conservative74
##6Green348
##7Labour749
##8Mana62
##9MaoriParty128
##10National1130
##11NZFirst216
##12UnitedFuture14

Because there is a %>% at the end of the line, R knows to continue on to the next line, as with any other to
be continued symbol at the end of the line.

Note that adding the lter also got rid of the NA entries. NA (Not Available) is used to indicate blank entries
those observations for which there is no data recorded. It is always a good plan to be aware of NAs and
deliberately include them in or exclude them from the analysis so that the nal results are not surprising. In
this case since NA indicates that these people did not answer the question about which party they voted for,
exluding them from the analysis makes sense.
We can also similarly view the levels and number of occurances of these levels in the X_singlefav variable:
selected_nzes2011%>%
group_by(X_singlefav)%>%
summarise(count=n())

###Atibble:82
##X_singlefavcount
##<chr><int>
##1Act33
##2Green388
##3Labour1043
##4Mana47
##5National1266
##6NZFirst138
##7UnitedFuture128
##8<NA>58

This set also has NA entries, but in this case we dont want to get rid of anything but the NA s so we need to
target them directly. NA entries need special targeting because they do not actually exist (they are dierent
to the text "NA" or a variable saved with the name NA ).
If we only wanted to nd the NA s we would use the is.na() function with the name of the variable inside
the parentheses.
However since we want the entries that are not NA s we can use the Not operator, ! , to indicate we want
all the ones that are not NA: !is.na() . Hence we can filter out all non NAs in our dplyr chain:
selected_nzes2011%>%
filter(!is.na(X_singlefav))%>%
group_by(X_singlefav)%>%
summarise(count=n())

###Atibble:72
##X_singlefavcount
##<chr><int>
##1Act33
##2Green388
##3Labour1043
##4Mana47
##5National1266
##6NZFirst138
##7UnitedFuture128

And remember that we can filter for multiple characteristics at once:

selected_nzes2011%>%
filter(!is.na(X_singlefav),jpartyvote!="Don'tknow")%>%
group_by(X_singlefav)%>%
summarise(count=n())

###Atibble:72
##X_singlefavcount
##<chr><int>
##1Act29
##2Green354
##3Labour914
##4Mana42
##5National1172
##6NZFirst119
##7UnitedFuture115

If we examine the categories in jdiffvoting we can see that this variable has levels such as both
"Don'tknow" and NA .
selected_nzes2011%>%
group_by(jdiffvoting)%>%
summarise(count=n())

###Atibble:72
##jdiffvotingcount
##<chr><int>
##1Don'tknow63
##2Votingcanmakeabigdifferencetowhathappens1605
##3Votingcanmakeareasonableamountofdifferencetowhathappens841
##4Votingcanmakesomedifferencetowhathappens339
##5Votingwon'tmakeanydifferencetowhathappens119
##6Votingwon'tmakemuchdifferencetowhathappens106
##7<NA>28

We need to decide how we want to handle these levels in our analysis.

Remember that our main question is about whether people vote for their favorite party or a dient one.
Hence an straighforwrd approach would be to rst determine whether each observation in the data
represents a person who voted for the party same as their favorite party or dierent. This requires creating
a new variable with the mutate() function.
In creating this variable we want to evaluate if for a given observation the values in the jpartyvote and
X_singlefav variables are the same, or dierent:
selected_nzes2011<selected_nzes2011%>%
mutate(sameparty=ifelse(jpartyvote==X_singlefav,"same","different"))

This creates a new variable named sameparty that has the value "same" if jpartyvote is equal to
X_singlefav , and "different" otherwise.
We can again check our work by exploring the groupings in a View:

selected_nzes2011%>%
group_by(jpartyvote,X_singlefav,sameparty)%>%
summarise(count=n())

##Source:localdataframe[82x4]
##Groups:jpartyvote,X_singlefav[?]
##
##jpartyvoteX_singlefavsamepartycount
##<chr><chr><chr><int>
##1ActActsame12
##2ActGreendifferent1
##3ActNationaldifferent14
##4ActUnitedFuturedifferent1
##5Act<NA><NA>1
##6ALCGreendifferent1
##7ALCLabourdifferent4
##8ALCNationaldifferent2
##9ALCUnitedFuturedifferent3
##10AllianceLabourdifferent1
###...with72morerows

We can see that observations where jpartyvote equaled X_singlefav , the value "same" was recorded for
the new variable sameparty , and the value "different" was recorded otherwise. If either jpartyvote or
X_singlefav had an NA , R could not check for equality and hence NA was recorded for the sameparty
variable as well.
To view and summarize the same entries we can use the following:
selected_nzes2011%>%
group_by(jpartyvote,X_singlefav,sameparty)%>%
summarise(count=n())%>%
filter(sameparty=="same")

##Source:localdataframe[7x4]
##Groups:jpartyvote,X_singlefav[7]
##
##jpartyvoteX_singlefavsamepartycount
##<chr><chr><chr><int>
##1ActActsame12
##2GreenGreensame237
##3LabourLaboursame632
##4ManaManasame31
##5NationalNationalsame1004
##6NZFirstNZFirstsame82
##7UnitedFutureUnitedFuturesame5

And to view and summarize the dierent entries we can use the following:

selected_nzes2011%>%
group_by(jpartyvote,X_singlefav,sameparty)%>%
summarise(count=n())%>%
filter(sameparty=="different")

##Source:localdataframe[59x4]
##Groups:jpartyvote,X_singlefav[59]
##
##jpartyvoteX_singlefavsamepartycount
##<chr><chr><chr><int>
##1ActGreendifferent1
##2ActNationaldifferent14
##3ActUnitedFuturedifferent1
##4ALCGreendifferent1
##5ALCLabourdifferent4
##6ALCNationaldifferent2
##7ALCUnitedFuturedifferent3
##8AllianceLabourdifferent1
##9AllianceNationaldifferent1
##10AnotherpartyGreendifferent2
###...with49morerows

We can also check how we got any NA s we have by using the is.na() function:
selected_nzes2011%>%
group_by(jpartyvote,X_singlefav,sameparty)%>%
summarise(count=n())%>%
filter(is.na(sameparty))

##Source:localdataframe[16x4]
##Groups:jpartyvote,X_singlefav[16]
##
##jpartyvoteX_singlefavsamepartycount
##<chr><chr><chr><int>
##1Act<NA><NA>1
##2Conservative<NA><NA>1
##3Don'tknow<NA><NA>7
##4Green<NA><NA>1
##5Labour<NA><NA>11
##6MaoriParty<NA><NA>2
##7National<NA><NA>7
##8NZFirst<NA><NA>2
##9<NA>Act<NA>4
##10<NA>Green<NA>32
##11<NA>Labour<NA>121
##12<NA>Mana<NA>4
##13<NA>National<NA>92
##14<NA>NZFirst<NA>17
##15<NA>UnitedFuture<NA>12
##16<NA><NA><NA>26

The checks show that the observations with NA s in the sameparty are going to be excluded from the analysis
when we ter out the NA s in the jpartyvote and X_singlefav variables, so we dont need to worry about
them anymore.

Step four. Prepare for the second question

As a second question, we might be interested in exploring the relationship between age of voters and how
much they like the NZ First party. We become familiar with the variables jnzflike and jage in the
codebook, then explore the data.
str(selected_nzes2011$jnzflike)

##Factorw/12levels"0","1","10","2",..:11410411NANA112...

str(selected_nzes2011$jage)

##int[1:3101]3737287143NA59686470...
jnzflike is a factor variable, in fact its ordinal and by default the levels are listed in alphabetical order.
Since this is a categorical variable, we can also summarize the occurances of each level with group_by() and
summarise() again:

selected_nzes2011%>%
group_by(jnzflike)%>%
summarise(count=n())

###Atibble:132
##jnzflikecount
##<fctr><int>
##10622
##21298
##310134
##42266
##53227
##64162
##75544
##86165
##97138
##108107
##11981
##12Don'tknow224
##13NA133

While jnzflike is on a 0 to 10 scale, this variable also has a level labeled "Don'tknow" , which is why R
stores this variable as not a numeric variable.
jage , on the other hand, is an integer, with values that are whole numbers between 0 and innity (or NA ).

For this variable we would want to take a look at numerical summaries such as means, medians, etc.

selected_nzes2011%>%
summarise(agemean=mean(jage),agemedian=median(jage),agesd=sd(jage),
agemin=min(jage),agemax=max(jage))

##agemeanagemedianagesdageminagemax
##1NANANaNNANA

What went wrong? The reason why all of the results were reported as NAs is that there were some NA
entries in the jage variable (people not reporting their age). Since it is not possible to take the average of a
series of values that contain NA s, obtaining the numerical summaries requires that we exclude the NA s
from the calculation.
Most numerical summary functions allow us to easily exclude NA s with the na.rm argument. See the help
documentation for the median function for more information.
?median

##startinghttpdhelpserver...

##done

An alternative approach is just to filter out the NA s rst, and then ask for the numerical summaries:
selected_nzes2011%>%
filter(!(is.na(jage)))%>%
summarise(agemean=mean(jage),agemedian=median(jage),agesd=sd(jage),
agemin=min(jage),agemax=max(jage))

##agemeanagemedianagesdageminagemax
##153.223285417.537118100

An age range of 18 to 100 is a reasonable age range for a voting age population, so there are no obvious
errors in the data. If there were, we would need to decide if we should lter them out of the analysis.
Having gained some familiarity with the specic variables we are using, we next need to consider if there is
additional work we should do on the data in investigating the question. There are a number of dierent
approaches we might take. For example, we could consider if those that strongly like NZ First are older than
those that strongly dislike NZ First, or we could consider if old people like NZ First more than young people.

Approach 1: Strongly liking and disliking NZ First and age

If we wanted to select only two of the possible levels in how much people like NZ First, we can lter for these
specic levels. When interested in ltering for multiple values a variable can take, the %in% operator can
come in handy:

selected_nzes2011%>%
filter(jnzflike%in%c("0","10"))%>%
group_by(jnzflike)%>%
summarise(count=n())

###Atibble:22
##jnzflikecount
##<fctr><int>
##10622
##210134

Remember that the jnzflike is not a numerical variable, hence we use the quotation marks around the
values (even though they happen to be numbers).
This is an example of simpligying the analysis by considering only two levels of a categorical variable, as
opposed to all possible levels.

Approach 2: Age and liking for NZ First

We might also like to rene our question slightly, asking do people above retirement age (65 in New Zealand)
like NZ First more than younger people. To do this we can turn the numeric age variable into a categorical
variable based on whether people are 65 years or older or younger than 65. Once again we make use of the
mutate() and ifelse() functions:
selected_nzes2011<selected_nzes2011%>%
mutate(retiredage=ifelse(jage>=65,"retiredage","workingage"))
selected_nzes2011%>%
group_by(retiredage)%>%
summarise(count=n())

###Atibble:32
##retiredagecount
##<chr><int>
##1retiredage876
##2workingage2156
##3<NA>69

We can see that individuals in the dataset are now labeled as either "retiredage" or "workingage" or
neither ( NA ), which we can easily lter out if need be.
This is an example of using a numerical threshold to convert a numerical variable to a categorical variable.
For approach 2, we might also be want to turn the scale of liking into numeric values, because at the
moment we cannot easily get summary information of the data in factor form. For example, if we ty to run
the following command, we get an error saying need numeric data.
selected_nzes2011%>%
group_by(retiredage)%>%
summarise(medlike=median(jnzflike))

it generates a need numeric data error.

We can change the type of data with functions of the form as.thingtochangeto() , but it is easy to go wrong
with factors. For example, this is wrong:
selected_nzes2011<selected_nzes2011%>%
mutate(numlikenzf=as.numeric(jnzflike))

We can see it has gone wrong if we use grouping to check our work (and it is a very good plan to check our
work after converting factors).
selected_nzes2011%>%
group_by(jnzflike,numlikenzf)%>%
summarise(count=n())

##Source:localdataframe[13x3]
##Groups:jnzflike[?]
##
##jnzflikenumlikenzfcount
##<fctr><dbl><int>
##101622
##212298
##3103134
##424266
##535227
##646162
##757544
##868165
##979138
##10810107
##1191181
##12Don'tknow12224
##13NANA133

Factor entries have two parts: the text we see on the screen, and a numeric order (remember how 10 was
coming between 1 and 2 because of the alphabetical order). When we say turn this into a number, R uses
the numeric order in which it stores the values to do that conversion, as opposed to the names of the levels
of the categorical variable. Hence, we need a conversion method that will use the text strings that label the
levels, as opposed to the storage order of these levels. We can do this by rst saving the variable as a
character variable, and then turning it into a number:
selected_nzes2011<selected_nzes2011%>%
mutate(numlikenzf=as.numeric(as.character(jnzflike)))

##Warningineval(substitute(expr),envir,enclos):NAsintroducedby
##coercion

The warning NAs introduced by coercion happens since the level "Don'tknow" cannot be turned into a
number. But this should be ne for our purposes since we are interested in the numerical responses
anyway.

Variable
selected_nzes2011%>% Question

DataType

group_by(jnzflike,numlikenzf)%>%
summarise(count=n())

##Source:localdataframe[13x3]
##Groups:jnzflike[?]
##
##jnzflikenumlikenzfcount
##<fctr><dbl><int>
##100622
##211298
##31010134
##422266
##533227
##644162
##755544
##866165
##977138
##1088107
##119981
##12Don'tknowNA224
##13NANA133

Converting the factor to a character rst ensures that the numerical values used in the labels of the levels of
the categorical variable are used.
Now that we cleaned up the data in a way that addresses the needs of the research questions we want to
explore, we are ready to continue with our analysis.

Appendix: List of elds in example data

Variable

Question

DataType

jactlike

A14: how much like Act

Factor

jactlr

A18: Act on left-right scale

chr

jage

Respondents age in years

int

jblogel

A6h: visit political blog for election

chr

jdiffvoting

A13: does voting make any dierence to what happens

chr

jdiscussp

A11a: how often discussed politics with others

chr

Jelect

Electorate

int

jelecvote

C4: if cast electorate vote, for which partys candidate

chr

jethnicity_a

F19a: ethnicity - Asian

chr

jethnicity_e

F19a: ethnicity - NZ European

chr

jethnicity_m

F19a: ethnicity - NZ Maori

chr

Variable

Question

DataType

jethnicity_o

F19a: ethnicity - Other

chr

jethnicity_p

F19a: ethnicity - Pacic

chr

jethnicityx

F19ax: other ethnic group belonged to detail

chr

jethnicmost

F19b: ethnic group identied with most

chr

jethnicmostx

F19bx: other ethnic group identied with most

chr

jfirstpx

C10x: on election day other party most wanted to be in government

chr

jgovpact

C8: Act helped form the government after 2008 election

chr

jgovpdk

C8: cant recall which parties formed the government after 2008
election

chr

jgovpgrn

C8: Greens helped form the government after 2008 election

chr

jgovplab

C8: Labour helped form the government after 2008 election

chr

jgovpmao

C8: Maori Party helped form the government after 2008 election

chr

jgovpmnp

C8: Mana Party helped form the government after 2008 election

chr

jgovpnat

C8: National helped form the government after 2008 election

chr

jgovpnzf

C8: NZ First helped form the government after 2008 election

chr

jgovunf

C8: United Future helped form the government after 2008 election

chr

jgrnlike

A14: how much like Greens

Factor

jgrnlr

A18: Greens on left-right scale

chr

jhhadults

F23a: number of adults in household

int

jhhchn

F23b: number of children in household

int

jhhincome

F22: household income between 1 April 2010 and 31 March 2011

chr

jhqual

F8: highest formal educational qualication

chr

jlablike

A14: how much like Labour

Factor

jlablr

A18: Labour on left-right scale

chr

jlanguage

F3: main language spoken at your home

chr

jlanguagex

F3x: other main language spoken

chr

jmaolike

A14: how much like Maori Party

Factor

jmaolr

A18: Maori Party on left-right scale

chr

jmarital

F24: marital status

chr

jmnplike

A14: how much like Mana Party

Factor

Variable

Question

DataType

jmnplr

A18: Mana Party on left-right scale

chr

jmostlike

A15: on election day which party liked most

chr

jmostlikex

A15x: other party liked most

chr

jnatlike

A14: how much like National

Factor

jnatlr

A18: National on left-right scale

chr

jnatradio

A10d: how often followed election news on Radio New Zealand:

National

chr

jnevervoteact

C16: would never vote for Act

chr

jnevervotegrn

C16: would never vote for Greens

chr

jnevervotelab

C16: would never vote for Labour

chr

jnevervotemao

C16: would never vote for Maori Party

chr

jnevervotemnp

C16: would never vote for Mana Party

chr

jnevervotenat

C16: would never vote for National

chr

jnevervoteno

C16: no party for which you would never vote

chr

jnevervotenzf

C16: would never vote for NZ First

chr

jnevervoteoth

C16: would never vote for another party

chr

jnevervoteothx

C16: other party for for which you would never vote

chr

jnevervoteunf

C16: would never vote for United Future

chr

jnewspaper

A10c: how often followed election news in newspaper

chr

jnzflike

A14: how much like NZ First

Factor

jnzflr

A18: NZ First on left-right scale

chr

jpartyvote

C3: if cast party vote, for which party

chr

jpcmoney

A11d: how often contributed money to a party or candidate

chr

jpcposter

A11e: how often put up party or candidate posters

chr

jpersuade

A11c: how often talk to anyone to persuade them how to vote

chr

jrallies

A11b: how often attended political meetings or rallies

chr

jrelang

F17: anglican

chr

jrelbap

F17: baptist

chr

jrelcath

F17: catholic

chr

jrelfun

F17: independent-fundamentalist-pentecostal church

chr

Variable

Question

DataType

jreligionx

F17x: other religion detail

chr

jreligiousity

F18: how religious are you

chr

jrellat

F17: latter day saints

chr

jrelmeth

F17: methodist

chr

jrelnonc

F17: non-Christian

chr

jrelnone

F17: no religion

chr

jrelothc

F17: other Christian

chr

jrelpres

F17: presbyterian

chr

jrelrat

F17: ratana

chr

jrelservices

F16: apart from weddings, funerals, baptisms, how often do you

attend religious services

chr

jrepublic

B1: should NZ become a republic or retain Queen as head of state

chr

jrollsex

Respondents gender from electoral roll

chr

jsecondp

C11: on election day which party overall was you second choice to be
in government

chr

jslflr

A19: yourself on left-right scale

chr

jspbusind

B3f: should there be more or less public spending on business and

industry

chr

jspdefence

B3d: should there be more or less public spending on defence

chr

jspedu

B3b: should there be more or less public spending on education

chr

jspenviro

B3i: should there be more or less public spending on the

environment

chr

jsphealth

B3a: should there be more or less public spending on health

chr

jsppolice

B3g: should there be more or less public spending on police and law
enforcement

chr

jspsuper

B3e: should there be more or less public spending on

superannuation

chr

jspunemp

B3c: should there be more or less public spending on unemployment

benets

chr

jspwelfare

B3h: should there be more or less public spending on welfare

benets

chr

jtalkback

A10e: how often followed election news on talkback radio

chr

junflike

A14: how much like United Future

Factor

Variable

Question

DataType

junflr

A18: United Future on left-right scale

chr

jwkdis

F9: disabled, unable to work

chr

jwkft

F9: working full-time for pay or other income

chr

jwkpt

F9: working part-time for pay or other income

chr

jwkret

F9: retired

chr

jwksch

F9: at school, university, or other educational institution

chr

jwkun

F9: unemployed, laid o, looking for work

chr

jwkunpi

F9: working unpaid within the home

chr

jwkunpo

F9: working unpaid outside the home

chr

njelecvote

Electorate Vote with nonvote

chr

njptyvote

Party Vote with nonvote

chr

r_jind

Respondent Industry Codes

chr

_singlefav

Caluclated Variable of most liked of major parties Question A14

chr