Writing a Dissertation See also: Writing a Research Proposal

The aim of the dissertation or thesis is to produce an original piece of research

work on a clearly defined topic.

Usually a dissertation is the most substantial piece of independent work in the

undergraduate programme, while a thesis is usually associated with master's
degrees, although these terms can be interchangeable and may vary between
countries and universities.

A dissertation or thesis is likely to be the longest and most difficult piece of

work a student has ever completed. It can, however, also be a very rewarding
piece of work since, unlike essays and other assignments, the student is able
to pick a topic of special interest and work on their own initiative.

Writing a dissertation requires a range of planning and research skills that will be of
great value in your future career and within organisations.

The dissertation topic and question should be sufficiently focused that you can
collect all the necessary data within a relatively short time-frame, usually about six
weeks for undergraduate programmes.

You should also choose a topic that you already know something about so that you
already have a frame of reference for your literature search and some understanding
and interest in the theory behind your topic.

There are many ways to write a dissertation or thesis.

Most universities and colleges provide very specific guidance to their students about
their preferred approach.

This page, and those that follow, are designed to give you some ideas about how you
might carry out your literature review, and then write each of the various sections of
your dissertation in the absence of, or in addition to, any specific guidance from your
university.
Organising your Time
However organised you are, writing your dissertation is likely to be one of the most
challenging tasks you have ever undertaken.

Take a look at our pages on Organising your Study Time and Organisation Skills, as well
as Project Management Skills and Project Planning, to give you some ideas about how to
organise your time and energy for the task ahead.

General Structure
Like an academic paper for journal publication, dissertations generally follow a fairly
standard structure. The following pages discuss each of these in turn, and give more
detailed advice about how to prepare and write each one:

Research Proposal
Introduction
Literature Review
Methodology
Results and Discussion
Conclusions and Extra Sections
Particularly for master's programmes, your university may ask for your thesis to be
submitted in separate sections, rather than as a single document. One breakdown that is
often seen is three-fold:

Introduction and/or Research Proposal, which should set out the research question
that you plan to explore and give some ideas about how you might go about it. If you are
submitting it as a research proposal, it will be fairly sketchy as you won’t have had a
chance to review the literature thoroughly, but it should contain at least some theoretical
foundation, and a reasonable idea of why you want to study this issue;
Literature Review and Methodology, which are often combined because what you
plan to do should emerge from and complement the previous literature; and
Results and Discussion, which should set out what you actually did, the results you
obtained, and discuss these in the context of the literature.

Warning!

You will probably have an overall word count for the total dissertation or thesis. If
you are required to submit in sections, ensure that you have left yourself enough
 words for the Results and Discussion. It is easy to get carried away with the
literature review.

As a general guide, use the marking scheme to show you the approximate split
for the word count. For example, if the introduction is worth 20%, and each of
the other two submissions 40%, for a total word count of 10,000 words, the
introduction should be at most 2,000 words, and each of the other two around
4,000 words.

If you’re submitting your dissertation as a single piece of work, and not in separate
submissions, you may find it easier not to write it in order.

It is often easier to start with the literature review and then write the methodology.

The introduction may be the last part you write, or you may wish to rewrite it once
you’ve finished to reflect the flow of your arguments as they developed.

Top Tip

One of the best ways to write a dissertation is as you go along, especially the
literature review.

As you read each reference, summarise it and group it by themes. Don’t forget
to reference it as you go!

You should be used to referencing by the time you write your dissertation but if
you need a refresher then see our page: Academic Referencing.

Writing Style
Dissertations and academic articles used always to be written in the third
person, and in the passive voice; as an example, you might write ‘An
experiment was carried out to test…’

However, many journals have now moved away from that convention and
request first person and active voice, which would require you to write ‘I
carried out an experiment to test…’

Check with your university about their requirements before you start to write.
If you cannot find any guidelines, then ask your supervisor and/or the person who will be
marking your thesis about their preferences. Make sure that the voice and person are
consistent throughout.

Whatever style is preferred, aim to keep your language simple and jargon-free. Use
shorter, simpler words and phrases wherever possible. Short sentences are good as they
are easier to follow. Any sentence that runs to more than three lines needs to be cut down
or split.

Phrases to avoid include:

Phrase Use instead

Due to the fact that… Because
In addition Additionally (or also)
In order to To
In the first place First
A considerable number Many
Whether or not Whether

Remember the Golden Rule

K.I.S.S. - Keep It Simple, Stupid

The Role of your Academic Supervisor

The role of your supervisor is to supervise your work. It is not to do it for you,
nor to tell you how to do it.

However, their academic reputation is bound up in the results of the students that
they supervise so they have a vested interest in helping you to get the best possible
marks. You should therefore not feel shy or embarrassed about asking them for help
if you get into difficulties, or if you need some advice.

Academics tend to take a highly personal approach to supervision. Some will be prepared
to spend a lot of time with you, talking about what you are planning to do by way of
research and your emerging findings. Others will have very little contact with you, apart
from being prepared to read a draft of your dissertation.

It’s worth spending a bit of time building up your relationship with your supervisor (have a
look at our page on Building Rapport for help). It’s also worth discussing and clarifying with
them exactly what they are prepared to do to support you, and in particular practical details
such as:

How often are they prepared to meet with you during your research?
How quickly will they respond to emails asking for advice and/or guidance?
How much time do they need to review drafts of work?
How many drafts of your work are they prepared to read? University guidelines usually
say ‘a first draft’ but many academics are prepared to read a preliminary draft to check
that you are on the right track, and then a more polished version.
Having reviewed a draft, will they send you comments by email, or would they prefer to
meet to discuss it?
One final piece of advice about your supervisor: if you don’t get on, then change supervisor.
But do so as early as possible. Nobody wants you or your supervisor to struggle with the
relationship, but they won’t be very sympathetic if you’re asking for a change a month
before your deadline.

Formatting and Templates

If your university has a required format for a dissertation, and particularly if they supply a
template, then use it! Start your writing straight into the template, or format your work
correctly from the start. There is very little worse than cutting and pasting your work
frantically into a template 10 minutes before your submission deadline. Templates are
designed to make your life easier, not harder.
You will also need to format the references in the university’s preferred style. It is easier to
do this as you go along. If the format is MLA, APA or Chicago, you can use Google Scholar
to format it for you: search for the article title, then click on ‘cite’. This will save you typing
out all the names, and can also be used, with minor tweaks, for other formats. But beware:
it’s not always right! If it looks odd, check the original source.

Proof-reading
You’ll need to give yourself plenty of time to proof-read your work, to make sure that you
haven’t made any stupid errors, and that it all flows correctly. This is likely to take longer
than you think. You’ll also need to do this when you’re fresh, not last thing at night when
you’re tired.

If possible, try to find a friend or fellow-student in the same position with whom you can
swap dissertations for proof-reading. Fresh eyes are likely to spot errors much more
effectively than those who already know what it should say.
Language Editing
The international language of academic publishing is English and many universities require
their students to publish their dissertations in English. If your first language is not English,
this is going to be a problem because your English will almost certainly not be up to the
task. You have two choices about how you approach this:

You find a native English speaker, perhaps a fellow student, who is prepared to read
your thesis for you and help you improve the English (preferably for free, or at least for
the price of not much more than a meal and a few drinks); or
You pay an editor to do the work for you. This will not be cheap; the going rate for
high quality academic language editing is about $7 per 250 words. You can find
professional language editors via the websites of publishers of academic journals such
as Emerald and Springer.
You will need to ensure that you build in sufficient time to allow someone else to read over
your work. Nobody, not even if you are paying them, is going to want to stay up all night to
edit your work because you left it too late. Many will also prefer not to work at weekends.
Allow at least two weeks for professional language editing.

A Note on Plagiarism

DO NOT PLAGIARISE

If you are found to have plagiarised you will be heavily penalised and will
probably lose your degree.

Ways to avoid being caught out inadvertently include:

Never copy and paste from a journal article. Always summarise it

in your own words, which also helps to make sure that you have
understood it.
If, for the sake of time, you want to copy and paste specific
sentences which sum up the argument particularly well, always put
them in quotation marks in your summary, with the source, so that
you will remember that they are direct quotes and need to be
acknowledged as such.
See our page: Academic Referencing for more information.

Conclusion
Research Methods See also: Writing a Research Proposal

Most people first encounter research as part of a school or college course.

A piece of research is usually included in any advanced degree course, and

may also be integral to undergraduate degrees. Basic research, such as
issuing questionnaires, may be undertaken in social science classes at school.

But there are many more applications for quality research.

These include market research to discover customer preferences, or to establish

whether a new product will sell, and focus groups to discuss politics.

Our Research Methods pages are designed to help you choose and then
use the right research method for your purposes.
They cover the whole process of research, from understanding the
philosophical theory underpinning your choice of method, through
choosing the methods that you will use to answer your research question,
to collecting data and then analysing it.

Introducing Research Methods

Your research method depends on the question that you wish to answer, and the
philosophy that underpins your view of research.

The best place to start is our page An Introduction to Research Methods. This sets out
the basic principles of research design, and the role of the researcher.

Our page on Designing Research explains how to approach research, and what to think
about in designing your research. It sets out some possible research approaches,
including experimental and quasi-experimental designs, survey research, and
ethnography.

Finally, you need to make a decision about whether your research will be Qualitative or
Quantitative, or even mixed.
 Qualitative Research Designs

Qualitative research is concerned with human behaviour, and why people act
the way that they do.

Common methods used for qualitative research include Interviews and Focus Groups
and Group Interviews. Both these methods allow researchers to explore a topic in depth
with one or two people at a time, or within a small group. You can also
collect Qualitative Data from Interactions, in research that recognises that the
researcher is a key part of the situation, rather than an outside observer.

Quantitative Research Designs

Quantitative research always collects numerical data.

If you are not collecting numbers, then your research is qualitative, not quantitative.
Quantitative research is usually used to get views from large numbers of people.

The first step in quantitative research is to determine your Sampling and Sample
Design. You then need to gather data. Suitable methods include surveys (and our page
on Surveys and Survey Design explains more about this surprisingly complex subject).
Other sources of data include Observational and Secondary Data.

Analysing Research Data

Your choice of analysis method will depend heavily on your choice of research
method.

For example, for qualitative research, you may need an approach like content
analysis, because you will have generated large amounts of data, often narrative in
form. Our page on Analysing Qualitative Data explains more.

Quantitative data is often analysed using statistical methods, which may be both
simple and more complex, depending on the question you are trying to answer. Our
Research Methods an Introduction See also: Writing a Research Proposal

You are most likely to have to carry out a piece of research as part of a course
of study, whether for an undergraduate or post-graduate degree.

However, there are also plenty of times when you may need to do some basic
research as part of a job or a voluntary role, whether it’s a simple survey to
find out what customers think or a more advanced piece of research.

This page introduces some basic principles of research design and discusses how
your view of the world affects your choice of methods and techniques.

The Basic Principles of Research Design

According to one of the most respected management research textbooks, written by
Easterby-Smith, Thorpe and Jackson, there are four main features of research
design, which are distinct, but closely related.

They are:

Ontology. How you, the researcher, view the world and the assumptions that you make
about the nature of the world and of reality.
Epistemology. The assumptions that you make about the best way of investigating the
world and about reality.
Methodology. The way that you group together your research techniques to make a
coherent picture.
Methods and techniques. What you actually do in order to collect your data and carry
out your investigations.
Easterby-Smith and colleagues liken these four to the rings of a tree trunk: the methods are
the outermost, and most visible, but without the inner ones, the outer one would die. All four
need to be coherent and consistent to create a viable research design.

These principles are the same, whether you are doing scientific research in a
laboratory or sending out a customer questionnaire.
Before choosing your methods, you need to understand how they fit with your ‘bigger
picture’ of the world, and how you choose to investigate it, to ensure that your work will be
coherent and effective.

The Underlying Philosophy

There are four main schools of ontology (how we construct reality), summarised in
this table.

Internal
Ontology Realism Relativism Nominalism
Realism
The world is The world is
Scientific laws Reality is
‘real’, and real, but it is
are basically entirely created
science almost
Summary created by by people, and
proceeds by impossible to
people to fit their there is no
examining and examine it
view of reality external ‘truth’
observing it directly
There is a single Truth exists, There are many There is no
Truth
truth but is obscure truths truth
Facts exist, and Facts are
Facts depend on Facts are all
can be revealed concrete, but
Facts the viewpoint of human
through cannot always
the observer creations
experiments be revealed
From: Management Research (4th edition), Easterby-Smith, Thorpe and Jackson

It will hopefully be clear that the underlying philosophy affects the choice of research
methods. For example, a realist will attempt to ‘uncover the truth’, whereas a relativist will
be interested in exploring different people’s ideas of the truth. The two will require quite
different approaches. However, none of these positions are absolutes. They are on a
continuum, with overlaps between them.

Within social sciences, there are also different epistemological approaches, or the
way in which you choose to investigate the world. The two main schools are
positivism and social constructionism.

Positivists believe that the best way to investigate the world is through objective
methods, such as observations. Positivism fits within a realist ontology.
Social constructionists believe that reality does not exist by itself. Instead, it is
constructed and given meaning by people. Their focus is therefore on feelings, beliefs
and thoughts, and how people communicate these. Social constructionism fits better
with a relativist ontology.
 All these philosophical approaches, both ontological and epistemological, are
valid. There are many eminent researchers working in all of these traditions and
schools, and many others who draw on multiple approaches depending on what
they are investigating. The important thing is that your research should be
internally consistent.
If you say that you are using a social constructionist approach within a
relativist ontology, your research will need to involve conversations.
Observing people ‘doing what they do’ will not deliver the results that you
need to answer your research questions.

Your chosen ontology and epistemology will have implications for your methodology.
Realists tend to use a positivist epistemology. They start with hypotheses. They gather facts
through experiments, with a view to proving or disproving their hypotheses, and therefore
confirming, or not, their theory. Clinical trials for new drugs or treatments are a good
example of realist/positivist research.

Relativists, on the other hand, tend to take a social constructionist view. They start with
questions. They use case studies and surveys to gather both words (views) and numbers,
which they use to triangulate and compare. From these, they generate theories.
Social constructionist approaches tend to draw on qualitative sources of data, and positivist
approaches on quantitative data.

Quantitative data is about quantities, and therefore numbers.

Qualitative data is about the nature of the thing investigated, and tends to be words
rather than numbers.

A Note on Data Sources

Primary data is gathered by the researcher themselves, whether through surveys,
interviews, or by counting atoms in a laboratory. Because it is collected for the purposes
of the study, it is intrinsically interesting, although the researcher will also need to make
some comment on it when publishing it.
Secondary data is published by someone else, usually a public body or company,
although it may also consist of archive material such as historical records. A researcher
using such data needs to generate new and original insights into it.
Researchers can either choose to use primary or secondary data for their studies. Both
have their advantages and disadvantages, and most researchers will use a combination of
the two.
Designing Research See also: Writing a Research Proposal

Our page Introduction to Research Methods explains that the philosophical

approach that you take to the world, and to its investigation, underpins the
methods that you use to carry out research.

This page explains some basic types of research, and their advantages and
disadvantages.

Your philosophy, and therefore your choice of research methods, is likely to be

influenced by many things: your colleagues’ views, your organisation’s approach,
your supervisor’s beliefs, and your own experience.

There is no right or wrong answer to choosing your research methods.

However, the method you choose needs to answer your research question.

For example:

If you want to explore the reasons why people choose certain careers, you are going
to need to talk to people. Counting the number of people choosing nursing will not tell
you why these choices are made.

If, on the other hand, you want to know whether more people opt for caring
professions, then you will want some hard data about applications to universities and
colleges, and job applications.

Approaching Research Five Questions:

Whatever approach you choose for your research, you need to consider five
questions:

1. What is the unit of analysis? For example, country, company or individual.

 2. Are you relying on universal theory or local knowledge? In other words, will your
results be generalisable, and produce universally applicable results, or are there local
factors that will affect your results?

3. Will theory or data come first? Should you read the literature first, and then develop
your theory, or will you gather your data and develop your theory from that? Recently,
opinion seems to have swung towards this being an iterative process.

4. Will your study be cross-sectional or longitudinal? Are you looking at one point in
time, or changes over time?
5. Will you verify or falsify a theory? You cannot conclusively prove any theory; the
best that you can do is find nothing that disproves it. It is therefore easier to formulate
a theory that you can try to disprove, because you only need one ‘wrong’ answer to do
so.

All Swans are White

One way of thinking about this is to formulate a theory that all swans are white.
To verify this, you would have to look at every swan in the world. In the UK, you
can gather a vast amount of data that suggests that your theory is correct, but
still not prove it conclusively.

To disprove it, in other words taking the falsification route, you simply have to
find one swan which is not white. Visit a zoo or Australia to find a black swan
and your theory can be discarded.

Some Basic Research Designs

There are several broad types of research design, some of which are broadly
quantitative, some qualitative, and some mixed.

Experimental Designs
These usually involve two groups: an experimental group, which receives an intervention of
some sort, and a control group, which either receives no intervention, or a non-effective
one. Clinical trials are usually of this type. The aim of research of this type is to remove all
possible alternative explanations for the results (high internal validity) and to make them as
generalisable as possible (high external validity).
 Useful when you want to test a particular intervention, and you can disguise whether it is being
used or not.
Less useful when you need to understand why something is happening.

Quasi-Experimental Designs
These are used when an experimental design would be ideal but is not possible, for
example because of the length of time required for the study or the difficulty of keeping an
experimental group separate from the control group. Researchers usually test before and
after an intervention to see what effect it has had. Again, these types of studies try to
maximise validity.

Useful when a full experimental design is not possible but you need that kind of separation of
the groups.
Less useful when you can carry out a full experimental design, or you need to understand why
something is happening.

Survey Research
See our page: Surveys and Survey Design for more information.
Surveys may be factual, inferential or exploratory. They may either start with an idea, and
try to prove it by collecting information, or collect a large amount of information and see
what emerges. The main issue with a survey is reliability: whether the survey accurately
assesses the desired variable. Surveys are usually pre-tested on a small sample before
being used more widely and, for this reason, many researchers choose to use established
questionnaires rather than develop their own whenever possible. They can be used either
on a large sample or as structured interviews on a smaller sample.

Useful when you want to gather data from lots of different people and you can formulate
questions that can be answered fairly simply.
Less useful when you want to explore individual experience in detail.

Action Research and Cooperative Inquiry

This type of research assumes that the researcher is a key part of the research, rather than
an external force. It emerged from the idea that the best way to learn about an organisation
is to try to change it, and also that those involved in change should be encouraged to
influence the change.

Useful when you, the researcher, are part of the organisation.

Less useful when you need to gather hard data from an objective viewpoint.

Ethnography
In this type of research, the researcher immerses him or herself in the research setting and
becomes a part of the group under study. Some of the original ethnographers went to live
with remote tribes in the jungle. Ethnographic studies are very authentic: the researcher
understands the organisation from the inside.

Useful for researchers who are internal to the organisation that they are studying, which often
happens for those on executive MBAs.
Less useful when you need an objective study.

Narrative approaches
These approaches gather information by developing or gathering stories about a particular
subject.

Useful for generating a ‘group history’ of events, or finding out about relationships or values.
Less useful when you need an objective approach.

Case studies
These take either one or several examples, and study it or them in detail, then draw out the
more general lessons for wider application. Researchers may try to take a more rigorous
approach to demonstrating validity, and ensure that logic is applied to any comparisons, or
focus on creating a detailed picture. Although a case study cannot prove a theory, it can be
used to disprove one if the data from the organisation do not fit the theory.

Useful when you want to find out about one organisation, when one organisation is considered
to be an exemplar, or to compare a few organisations and identify the key differences in
approach.
Less useful for drawing generalised lessons that can be applied to any other organisation,
although there may be some.

Grounded theory
This approach examines the same event or process in several different settings or
organisations. The researcher carries out a process of sampling, making comparisons
between samples, using these to evolve a theory. When no new insights emerge from new
data, the researcher has reached theoretical saturation. Some experts recommend no prior
reading, but others suggest reading the literature beforehand to familiarise yourself with the
territory.

Useful when you have plenty to time to immerse yourself and do repeat samplings, testing your
data on subsequent samples.
Less useful when you need to do something quickly and produce results immediately.
Research methods are split broadly into quantitative and qualitative methods.

Which you choose will depend on your research questions, your underlying
philosophy of research, and your preferences and skills.

Our pages Introduction to Research Methods and Designing Research set out
some of the issues about the underlying philosophy.

This page provides an introduction to the broad principles of qualitative and

quantitative research methods, and the advantages and disadvantages of each in
particular situations.

Some definitions

Quantitative research is “explaining phenomena by collecting numerical

data that are analysed using mathematically based methods (in particular
statistics).”*

Qualitative research seeks to answer questions about why and how

people behave in the way that they do. It provides in-depth information
about human behaviour.

* Taken from: Aliaga and Gunderson ‘Interactive Statistics ‘3rd Edition (2005)

Quantitative Research
Quantitative research is perhaps the simpler to define and identify.
The data produced are always numerical, and they are analysed using mathematical
and statistical methods. If there are no numbers involved, then it’s not quantitative
research.

Some phenomena obviously lend themselves to quantitative analysis because they are
already available as numbers. Examples include changes in achievement at various
stages of education, or the increase in number of senior managers holding management
degrees. However, even phenomena that are not obviously numerical in nature can be
examined using quantitative methods.

Example: turning opinions into numbers

 If you wish to carry out statistical analysis of the opinions of a group of
people about a particular issue or element of their lives, you can ask
them to express their relative agreement with statements and answer on
a five- or seven-point scale, where 1 is strongly disagree, 2 is disagree, 3
is neutral, 4 is agree and 5 is strongly agree (the seven-point scale also
has slightly agree/disagree).

Such scales are called Likert scales, and enable statements of opinion
to be directly translated into numerical data.

The development of Likert scales and similar techniques mean that most
phenomena can be studied using quantitative techniques.

This is particularly useful if you are in an environment where numbers are highly valued
and numerical data is considered the ‘gold standard’.
However, it is important to note that quantitative methods are not necessarily the most
suitable methods for investigation. They are unlikely to be very helpful when you want to
understand the detailed reasons for particular behaviour in depth. It is also possible that
assigning numbers to fairly abstract constructs such as personal opinions risks making
them spuriously precise.

Sources of Quantitative Data

The most common sources of quantitative data include:

Surveys, whether conducted online, by phone or in person. These rely on the same
questions being asked in the same way to a large number of people;
Observations, which may either involve counting the number of times that a particular
phenomenon occurs, such as how often a particular word is used in interviews, or
coding observational data to translate it into numbers; and
Secondary data, such as company accounts.

Our pages on Survey Design and Observational Research provide more information about
these techniques.

Analysing Quantitative Data

There are a wide range of statistical techniques available to analyse quantitative data, from
simple graphs to show the data through tests of correlations between two or more items, to
statistical significance. Other techniques include cluster analysis, useful for identifying
relationships between groups of subjects where there is no obvious hypothesis, and
hypothesis testing, to identify whether there are genuine differences between groups.
 Our page Statistical Analysis provides more information about some of the simpler statistical
techniques.

Qualitative Research
Qualitative research is any which does not involve numbers or numerical data.

It often involves words or language, but may also use pictures or photographs and
observations.

Almost any phenomenon can be examined in a qualitative way, and it is often the preferred
method of investigation in the UK and the rest of Europe; US studies tend to use
quantitative methods, although this distinction is by no means absolute.

Qualitative analysis results in rich data that gives an in-depth picture and it is particularly
useful for exploring how and why things have happened.

However, there are some pitfalls to qualitative research, such as:

If respondents do not see a value for them in the research, they may provide
inaccurate or false information. They may also say what they think the researcher
wishes to hear. Qualitative researchers therefore need to take the time to build
relationships with their research subjects and always be aware of this potential.
Although ethics are an issue for any type of research, there may be particular
difficulties with qualitative research because the researcher may be party to
confidential information. It is important always to bear in mind that you must do no
harm to your research subjects.
It is generally harder for qualitative researchers to remain apart from their
work.By the nature of their study, they are involved with people. It is therefore helpful to
develop habits of reflecting on your part in the work and how this may affect the
research. See our page on Reflective Practice for more.

Sources of Qualitative Data

Although qualitative data is much more general than quantitative, there are still a
number of common techniques for gathering it. These include:

Interviews, which may be structured, semi-structured or unstructured;

Focus groups, which involve multiple participants discussing an issue;
‘Postcards’, or small-scale written questionnaires that ask, for example, three or four
focused questions of participants but allow them space to write in their own words;
Secondary data, including diaries, written accounts of past events, and company
reports; and
 Observations, which may be on site, or under ‘laboratory conditions’, for example,
where participants are asked to role-play a situation to show what they might do.

Our pages on Interviews for Research, Focus Groups and Observational Researchprovide
more information about these techniques.

Analysing Qualitative Data

Because qualitative data are drawn from a wide variety of sources, they can be
radically different in scope.

There are, therefore, a wide variety of methods for analysing them, many of which involve
structuring and coding the data into groups and themes. There are also a variety of
computer packages to support qualitative data analysis. The best way to work out which
ones are right for your research is to discuss it with academic colleagues and your
supervisor.

Our page Analysing Qualitative Data provides more information about some of the most
common methods.

It’s your research…

Finally, it is important to say that there is no right and wrong answer to which
methods you choose.

Sometimes you may wish to use one single method, whether quantitative or
qualitative, and sometimes you may want to use several, whether all one type or a
mixture. It is your research and only you can decide which methods will suit both
your research questions and your skills, even though you may wish to seek advice
from others.
Interviews for Research See also: Focus Groups and Group Interviews

Interviews are one way to gather data for research. The data gathered are
usually, though not always, qualitative in nature.

Interviews are usually used to explore a topic or topics in considerable depth

with a few people. They are not, in general, very useful for eliciting answers to
straightforward questions from large numbers of people, when a questionnaire
may be more useful.

Although interviews are useful for eliciting in-depth information, they do need careful
planning. Before you start, you need to be very clear what areas you want to explore,
and that an interview is the best way to do this. In general, interviews are most useful
when you wish to discover someone’s viewpoint and why they hold that view,
especially when the information is likely to be sensitive.

For more about alternative research methods, see our page Qualitative and Quantitative
Research and for more general interviewing skills see Interviewing Skills.

Level of Structure
The first issue to decide is how structured your interview should be.

Highly structured interviews are often used in market research and are basically
guided questionnaires (see our page on Surveys and Survey Design for more).

A more usual form, especially for management research and social sciences, is the
semi-structured interview.

In semi-structured interviews, the interviewer starts with a list of general introductory

questions or topics that they wish to explore. These questions will be used as a starting
point for discussion with all the interview subjects, but the format allows you to explore
interesting areas in more detail. This format is particularly useful in three cases:
When your interviewees are likely to provide you with interesting data in different areas
but you’re not sure exactly who will provide which information;
 When you want to be able to discuss emerging findings with your interviewees and test
out ideas with them, without being held within a rigid structure; and
When your interviewee is nervous about what you might want to discuss, because you
can send the outline of questions to them in advance.

Warning!

You can also use unstructured interviews, a more ethnographic technique,

aimed at exploring your interviewee’s knowledge. However, in research
designed to explore a specific issue, rather than just find out what someone
knows, this is likely to lead to confusion in the interviewee’s mind about what
you want to explore.

Face-to-Face or Telephone?
One-to-one interviews are usually held either face-to-face or on the phone. Both
techniques have advantages and disadvantages.

Face-to-face Telephone
Offer the richest data in terms of
Often easier to arrange, and
body language and non-verbal
Advantages managers may prefer them
communication as well as what is
because they are less formal
actually said.
Easier for interviewees to
Can be hard to arrange, especially cancel at short notice.
with busy people.
Disadvantages Difficult to conduct
It takes time and money to travel to effectively if you do not
meet interviewees. already have a relationship
with the other person.

As with any research, whether you choose face-to-face or telephone interviewing is often
a matter of compromise on time, money, and convenience, in which quality of interaction
may suffer. However, with care it is possible to get very good results from telephone
interviewing, especially when you already have a relationship with the interviewee.

Skills Needed for Effective Interviewing

As in any personal interaction, you will get a better quality interaction if you
take a bit of time to build rapport with your interviewee.

Of course, if you already have an established relationship, this won’t be a

problem, but this is often not the case in research, and five minutes at the start
of the interview can pay dividends later on in the quality of the data that you
obtain.

In qualitative research, with semi-structured interviews, the way that you ask the
questions is much less likely to lead to bias than in straightforward surveys.
However, there is still a danger of bias if you are tempted consciously or
subconsciously to impose your frame of reference onto your interviewee. To avoid
this:

Try to use open questions whenever possible as they are least likely to bias answers
(see our pages Questioning Skills and Techniques and Types of Questions for more
information); and
Use techniques of reflection and other clarification techniques to ensure that you have
fully understood your interviewee’s meaning, and also to prompt them to say more.

Warning!

To avoid bias, never use leading questions in an interview.

Interviewers may, however, need to probe deeper into a subject and, for this, specific
questioning techniques can be useful.

Types of Probe

The basic probe is repeating the initial question, which reminds the
interviewee what you asked. This is useful if they have wandered
off the subject.
Explanatory probes are questions like ‘What did you mean by
that?’ and ‘What makes you say that?’ and are useful for exploring
meaning further.
Focused probes include questions like ‘What sort of…?’
 The silent probe is where the interviewer simply remains silent and
waits for the interviewee to say more.
Drawing out is useful when the interviewee seems to have stopped
mid-sentence or mid-idea. Repeat the last few words that they said
with an upward inflexion, like a question, or add ‘Tell me more
about that’.
Giving ideas or suggestions would use questions like ‘Have you
thought about x?’ or ‘Have you tried…?’

Source: Easterby-Smith, Jackson and Thorpe (2012) Management Research.

Laddering
Laddering is a very specific interviewing technique which asks ‘Why?’-type
questions repeatedly to explore the interviewee’s values and motivations.

This is a powerful technique, but you need to be aware that people may be uncomfortable
with this until you have developed at least a superficial rapport and relationship. You can
also ‘ladder’ in the opposite direction, where you get more specific until you reach
examples, by asking questions like ‘Can you give me a specific example of that?’ or ‘When
was the last time that you remember something like that happening?’.
Perhaps most crucially, interviewers need to have very good listening skills. They need to
listen to what their interviewee is saying, and also be aware of what they are not saying, but
without imposing their own views.

Finally…
Interviews are a very good way of gathering rich qualitative information from a
limited number of people.

While you need enough data to make the research worthwhile, don’t try to interview
too many people. The quality lies in the depth of exploration, not necessarily in the
breadth of views.

If you need breadth, try a questionnaire or a focus group instead.

Focus Groups and Group Interviews
See also: Qualitative Data from Interactions

In most research situations, an ‘interview’ is a one-to-one interaction and our

page Interviews for Research discusses their use in gathering data for
research. However, there are times when a group interview, also known as a
focus group, can be the best way to ensure that you gain the range of views
that you need.

Focus groups are widely used in market research and in politics, but perhaps
less often in research situations. This may be because researchers lack the
necessary skills to make them work, but also because they are not sure when
such techniques would be most useful.

Using Focus Groups

Examples of when you might use a focus group are:

When you are short of time and you need to gather views from a
group of people, not just one or two individuals;
To review a process or event and gather different opinions about it
and how to improve it in future;
When opinions are not likely to be sensitive, and the subject is one
that can be freely discussed in a group without embarrassment or
concerns;
When you know that there is a range of views;
When you want to gather reactions from several people to an
event, especially as it happens. This is often used in politics, for
example to find out what people think about a party conference
speech, party political broadcast, or debate, where events are
often televised live.
Organising a Focus Group: Some Practical Thoughts
There are a number of issues to consider in running a focus group.

First, the practicalities. As with any research, select your sample carefully (see our
page on Sample and Sample Design for more). Focus groups are necessarily small,
which makes it much harder to use a representative sample. However, any sample
which is obviously not representative will risk invalidating your research.

Secondly, consider the venue and timing. It sounds obvious, but if you want to chat to
working people, you may want to hold the focus group at lunchtime. Nobody’s going to want
to stay late after work to talk to you. However, if you want to talk to mums of young children,
evening might be best, when their partners are home and the baby is in bed. Alternatively, if
you need to hold the event during the day, you might need to consider providing a crèche.
Consideration of these aspects avoids eliminating whole sections of your sample
population.

You also need to think about the location: is it accessible by public transport? If you’re
holding an evening event, will people feel safe walking there? And getting home? Will you
pay for travel costs? Will you provide a meal, light refreshments, or nothing at all? Ideally,
you want them to feel comfortable and relaxed, but will you achieve this best with the group
sitting round a table, like a formal business meeting, or in comfortable chairs, as if they
were chatting to friends?

All these will affect the ‘feel’ of the focus group and therefore the participants’ willingness to
contribute. The important thing to remember is that you can’t necessarily predict how.
You just need to be aware of it and take what steps you can to avoid any problems.
However, if any issues do occur, you’ll have to react to them on the day.

Finally, consider how you are going to record the event. Will you take notes? Will the group
record views on a flip chart? Or will you video/tape the whole event and review it later? You
will need informed consent from the group for whichever method you choose.

On the Day: Running a Focus Group

The most important thing to remember is that you are not trying to interview each
person individually. The idea is to create an environment in which each person feels
comfortable expressing their views when they wish to do so, and the group is able to
discuss the issues. This means one in which conversation flows, and no one
individual is allowed to dominate.

Like semi-structured interviews, focus groups will need a broad structure, including some
starting questions. If you wish to explore several different areas, make sure that you
manage the discussion to cover them all. This means that you may need to move the
conversation on from an area of interest to the participants to one that is more interesting to
you, but without alienating anyone. You’ll also want to have room for the discussion to
expand into areas that develop in the course of the event.

See our page: Facilitation Skills, for more about facilitating groups and some
useful techniques for managing them.

Getting Help...

If you’re anxious about running a focus group, you may find it helpful to talk to
your supervisor or your sponsor to discuss whether you can draft in some expert
help as there are plenty of consultancies who can provide this service at
reasonable prices.

Possible Problems
There are a number of criticisms that can be made of focus groups. These include:

Concerns that people may not feel comfortable airing their views in public. This
may be made worse both if they do not know the other participants and, paradoxically, if
they do. For example, if one participant is senior to another in the same organisation, the
more junior person may feel unable to express different views, but greater trust between
participants usually leads to more openness.
In public, people often express the views that they feel that they ought to have,
rather than their real views. This is known as social pressure and may mean that the
focus group’s views are more or less extreme in reality than the expressed view.
However, these concerns can be overcome by good facilitation at the event, including
careful design and outline of the topics to be covered, together with triangulation of your
research using other techniques and research methods.

A Concluding Thought
Focus groups are not ideal for every situation or every piece of research, and there
are serious questions to address in designing one. However, there is no question
that, if used well, focus groups are a strong tool for researchers to explore diverse
views, especially if time is too tight to allow sufficient in-depth one-to-one interviews
to be held. They are well worth exploring if you have the necessary facilitation skills,
or access to support to run such an event.
Qualitative Data from Interactions
See also: Focus Groups and Group Interviews

Our page on Analysing Qualitative Data discusses techniques for analysing

language-based data. This page explains more about how to generate,
analyse and present data identified through interaction with your research
subjects.

Here, the analysis is almost part of the process of generating the data, rather
than something done later on data gathered in a separate process.

In particular, this page discusses two techniques, repertory grid

analysis and cognitive mapping, which may be helpful for identifying individuals’
views of the world.

Repertory Grid Analysis

A repertory grid allows you to represent someone’s view of the world.

Repertory grids are particularly useful for investigating ideas that people may not have
thought about in any detail, or which are hard to articulate, as it puts a framework around
the discussion. It’s also based on the interviewees’ own constructs, and not those of the
researcher, which means that it avoids the researcher imposing their views on the research
subject.
However, creating a grid can take quite a long time, so you need willing interviewees. It also
requires considerable skill from the interviewer to explore all the constructs adequately.

See our page on Interviews for Research for more about interviewing skills.

Repertory Grid Analysis Step-by-Step

1. Decide on the focus of the grid, which should be quite specific.

Examples include the qualities that make a good manager or even
 what makes the ideal kitchen or similar.

2. With the interviewees, select a group of between five and ten

elements that are both relevant and will provide a good range. For
example, for the features that make a good manager, the elements
should be managers known to the interviewees, some of whom they
regard as good and some of whom are not so good, with a range in
between.

3. Write each element onto a separate card.

4. Pick three at a time, and ask the interviewees to identify the ‘odd one
out’, and provide one word that describes the two that are similar,
and another for the one that is different. These two words become a
construct on a continuum between the two.
5. Continue selecting trios of elements until you have between six and
ten constructs.

6. Rate each of the elements against each of the constructs. You can
either just allocate them to one end or the other, or you can use a
five-point scale in between the two.

You can analyse small grids by eye, looking for relationships between elements and
constructs.

Larger grids often need computer assistance for analysis. The idea is to group the
constructs together in some way to form two or three bigger constructs which explain the
variation. At this stage, it may be clear why you’d need a computer programme to help with
the analysis of a large grid! Examples of suitable programmes include GridSuite.

You can either represent your findings as:

A map or graph, with two axes representing the two main ‘grouped constructs’ that
explain the most variation; or
A dendrogram, where the distance between ‘branches’ shows the relationship between
the constructs.
These visual representations are very helpful in showing a picture of the findings. Your final
step in any analysis is to reflect your findings back to your interviewees. This may well
produce new insights from them into the constructs or elements.

Warning!

Repertory grids are not easy to prepare or analyse. They take considerable time
and skill in interviewing to draw out all the aspects of each construct. Care is
 also needed to ensure that the results are meaningful to the interviewees and to
the researcher.

Cognitive Mapping
Cognitive maps are a little like mind maps. They are a visual representation
of an individual or group’s way of looking at the world, including values,
beliefs, attitudes and ideas, and how they relate to each other. It’s very much a
subjective view, and can only be drawn up with the full cooperation of the
interviewee(s).

Individual Maps
Individual maps are drawn up with just one person. As with repertory grid analysis, it helps
to have good interviewing skills, including strong questioning and listening skills.

By using the technique of laddering, described on our page Interviewing for Research, you
can explore values and beliefs as well as experience. As you elicit ideas and values, you
draw a map linking them to others.

It’s good to keep reflecting back to your interviewee to ensure that they agree with your
suggested links.

Dominoes
Dominoes is a specific technique used to explore values and beliefs when there may be
issues of power or control.

The researcher identifies elements, which may be people or objects, and places cards with
one on each in front of the interviewee, all at the same time. The interviewee is asked to
group them according to patterns, and explain their thinking.

They should reform the groups in as many ways as they can. Their comments are recorded
by the interviewer, who can also ask questions to clarify where necessary. This technique
can speed up the process of identifying constructs.

Group Maps
Although not used so often in pure research, group maps are often used by facilitators to
help groups to think through strategic issues. They can also be used as part of mediationto
help groups to understand each other’s points of view. They are a useful way of making a
visual representation of values, concerns and issues, and, as with many visual techniques,
are easy to understand quickly.

This makes group maps ideal for strategic planning processes (see our pages: Strategic
Thinking and Creative Thinking for more ideas of other techniques to use).

Conclusion
Both repertory grid analysis and cognitive mapping produce visual representations
of quite complex situations.

They are therefore helpful in simplifying and visualising ideas. However, it’s important to
remember that both require skilful interviewing and interaction with people to obtain the best
results. That said, even a fairly unskilled approach will produce some usable results, and
it’s always worth having a go at a new technique.

To get maximum value, however, it is worth trying to find someone else with more
experience to help you, at least the first time.
Sampling and Sample Design Part of our: Research Methods library.

When you collect any sort of data, especially quantitative data, whether
observational, through surveys or from secondary data, you need to decide
which data to collect and from whom.

This is called the sample.

There are a variety of ways to select your sample, and to make sure that it
gives you results that will be reliable and credible.

The difference between population and sample

Ideally, research would collect information from every single member of

the population that you are studying. However, most of the time that would
take too long and so you have to select a suitable sample: a subset of the
population.

Principles Behind Choosing a Sample

The idea behind selecting a sample is to be able to generalise your findings to the
whole population, which means that your sample must be:

Representative of the population. In other words, it should contain similar proportions of

subgroups as the whole population, and not exclude any particular groups, either by
method of sampling or by design, or by who chooses to respond.
Large enough to give you enough information to avoid errors. It does not need to
be a specific proportion of your population, but it does need to be at least a certain size
so that you know that your answers are likely to be broadly correct.
If your sample is not representative, you can introduce bias into the study. If it is not large
enough, the study will be imprecise.
However, if you get the relationship between sample and population right, then you
can draw strong conclusions about the nature of the population.

Sample size: how long is a piece of string?

How large should your sample be? It depends how precise you want the
answer. Larger samples generally give more precise answers.

Your desired sample size depends on what you are measuring and the size of
the error that you’re prepared to accept. For example:

To estimate a proportion in a population:

Sample size =[ (z-score)² × p(1-p) ] ÷ (margin of error)²

The margin of error is what you are prepared to accept (usually between
1% and 10%);
The z-score, also called the z value, is found from statistical tables and
depends on the confidence interval chosen (90%, 95% and 99% are
commonly used, so choose which one you want);
p is your estimate of what the proportion is likely to be. You can often
estimate p from previous research, but if you can’t do that then use 0.5.

To estimate a population mean:

Margin of error = t × (s ÷ square root of the sample size).

Margin of error is what you are prepared to accept (usually between 1%

and 10%);;
As long as the sample size is larger than about 30, t is equivalent to the z
score, and available from statistical tables as before;
s is the standard deviation, which is usually guessed, based on previous
experience or other research.

If you’re not very confident about this kind of thing, then the best way to deal
with it is to find a friendly statistician and ask for some help. Most of them will be
delighted to help you make sense of their specialty.

It is better to be imprecisely right than precisely wrong.

 How bias and precision interact:

Bias
High Low
High Precisely wrong Precisely right
Precision
Low Imprecisely wrong Imprecisely right
Source: Management Research (4th Edition), Easterby-Smith, Thorpe and Jackson

Imprecisely right means that you know broadly what the correct answer is. Precisely wrong
means that you think you know the answer, but you don’t. In other words, if you can only
worry about one, worry about bias.

Selecting a Sample
Probability sampling is where the probability of each person or thing being part
of the sample is known. Non-probability sampling is where it is not.

Probability Sampling
Probability sampling methods allow the researcher to be precise about the
relationship between the sample and the population.

This means that you can be absolutely confident about whether your sample is
representative or not, and you can also put a number on how certain you are about your
findings (this number is called the significance, and is discussed further in our page
on Statistical Analysis).

In simple random sampling, every member of the population has an equal chance of
being chosen. The drawback is that the sample may not be genuinely representative. Small
but important sub-sections of the population may not be included.

Researchers therefore developed an alternative method called stratified random

sampling. This method divides the population into smaller homogeneous groups, called
strata, and then takes a random sample from each stratum.

Proportional stratified random sampling takes the same proportion from each stratum,
but again suffers from the disadvantage that rare groups will be badly represented. Non-
proportional stratified sampling therefore takes a larger sample from the smaller strata, to
ensure that there is a large enough sample from each stratum.
Systematic random sampling relies on having a list of the population, which should
ideally be randomly ordered. The researcher then takes every nth name from the list.
 Warning!

There are many different methods of selecting ‘random samples’. If you are the
lead researcher for a project and instructing others to ‘take a random sample’, or
indeed asked to take a ‘random sample’, make sure you are all using the same
method!

Cluster sampling is designed to address problems of a widespread geographical

population. Random sampling from a large population is likely to lead to high costs of
access. This can be overcome by dividing the population into clusters, selecting only two or
three clusters, and sampling from within those. For example, if you wished to find out about
the use of transport in urban areas in the UK, you could randomly select just two or three
cities, and then sample fully from within these.

It is, of course, possible to combine all these in several stages, which is often done for
large-scale studies.

Non-Probability Sampling
Using non-probability sampling methods, it is not possible to say what is the
probability of any particular member of the population being sampled. Although this
does not make the sample ‘bad’, researchers using such samples cannot be as
confident in drawing conclusions about the whole population.

Convenience sampling selects a sample on the basis of how easy it is to access. Such
samples are extremely easy to organise, but there is no way to guarantee whether they are
representative.

Quota sampling divides the population into categories, and then selects from within
categories until a sample of the chosen size is obtained within that category. Some market
research is this type, which is why researchers often ask for your age: they are checking
whether you will help them meet their quotas for particular age groups.

Purposive sampling is where the researcher only approaches people who meet certain
criteria, and then checks whether they meet other criteria. Again, market researchers out
and about with clipboards often use this approach: for example, if they are looking to
examine the shopping habits of men aged between 20 and 40, they would only approach
men, and then ask their age.

Snowball sampling is where the researcher starts with one person who meets their
criteria, and then uses that person to identify others. This works well when your sample has
very specific criteria: for example, if you want to talk to workers with a particular set of
responsibilities, you might approach one person with that set, and ask them to introduce
you to others.

WARNING!

Non-probability sampling methods have generally been developed to address

very specific problems. For example, snowball sampling deals with hard-to-find
populations, and convenience sampling allows for speed and ease.

However, although some non-probability sampling methods, particularly quota

and purposive sampling, ensure the sample draws from all categories in the
population, samples taken using these methods may not be representative.

A Word in Conclusion
Almost all research is a compromise between the ideal and the possible.

Ideally, you would study the whole population; in practice, you don’t have time
or capacity. But care in your sample selection, both size and method, will
ensure that your research does not fall into the traps of either introducing bias,
or lacking precision. This, in turn, will give it that vital credibility.
Surveys and Survey Design See also: Designing Research

Surveys, which are also called questionnaires, are one of the key ways to
gather quantitative data for analysis.

Surveys rely on asking the same question in the same way to a large number
of people, and obtaining a lot of responses. These responses are then
analysed using statistical techniques to obtain information that can be
generalised about the whole population.

This page does not discuss semi-structured questionnaires often used as interview
outlines for qualitative research, but you can find out more about these in our
pages Interviews for Research and Focus Groups.

You can also find out more about identifying a suitable sample in our page
on Sampling and Sample Design.

Types of Survey
There are two main types of survey, self-completed and interviewer-administeredsurveys:

Self-Completed Surveys
A self-completed questionnaire is, as the name suggests, completed by the survey
subject. This may be either a postal questionnaire, or a web-based survey. Both of
these are much cheaper than face-to-face interviews.

Postal questionnaires are the old-fashioned way, but they do ensure that you do not
eliminate a large section of the population because they don’t have computers or internet
access. On the other hand, they often have a low response rate because if people don’t
complete them immediately they tend to forget about them. There is also no guarantee
whether the addressee has completed the survey themselves and so it’s very important to
check the accuracy and completeness of the answers before using the data for analysis.

There are now many tools available for web-based surveys, such
as SurveyMonkey and Qualtrics. These make administering a survey quicker and easier, as
well as cheaper, because there are no postal costs involved. However, there may be other
costs involved: for example, SurveyMonkey offers a membership level allowing free surveys
but they can have a maximum of 10 questions.

Web-based surveys can also improve response rates and completeness by being able
to use pop-up boxes to explain difficult sections, or to prompt for missing data, and they can
be customised for individuals much more easily than a postal survey. It’s also helpful to
have the data in electronic form without the need for any data input before analysis.

SkillsYouNeed uses LimeSurvey an open source survey tool - take either our Interpersonal
Skills Self-Assessment or What Sort of Leader are You? quiz to see this in action.

Interviewer-Administered Surveys
The alternative to self-completion is interviewer-administered surveys. These can be
either face-to-face or telephone interviews. Both have their drawbacks.

Face-to-face surveys are expensive and time-consuming. However, they do allow the
interviewer to check the respondent’s understanding and answer before completion, which
improves the accuracy of the data, and can also mean that hard-to-reach groups can be
included in the survey population. For example, those without access to technology are
often older, which would introduce bias into the sample if you used a web-based survey.

Telephone surveys combine the best of both worlds: an ability to interact with the
respondent without the expense of travelling. This means that the data can be accurate and
complete, but is obtained more cheaply than with face-to-face interviews. However, some
people object to telephone interviews, especially without prior arrangement. It is therefore
probably best used for management research, where you can set up the interview in
advance. There are, and have been, large scale studies that have used telephone
interviews. However, it is important to consider your sample selection to avoid any bias,
such as exclusion of the mobile-only population, or those who need help to have a
telephone conversation.

For more about sample selection, see our page on Sampling and Sample Design.

Principles of Survey and Questionnaire Design

It is very easy to write poor survey questions. Good ones take a bit more
thought and effort, but the good news is that there are five fairly
straightforward principles of good survey design.

1. Each question or item should only express one idea.

It’s very easy to overcomplicate. Pare your questions back to make sure that they only
cover one idea. If necessary, turn one question into two.

Example of a bad question: Do you agree that food prices are increasing to the point of
unaffordability?

Someone might agree that food prices are increasing, but not to the point of unaffordability.
You won’t know. Instead, make it two questions:

Do you agree that food prices are increasing?

How much do you agree that they will soon be unaffordable?

2. Avoid jargon, abbreviations and colloquialisms

3. Use simple language and expressions

Points two and three are similar, but slightly different. Whenever you write anything, you
should make your meaning as plain as possible. But this is even more important in surveys.

You want to be confident that your respondents have understood the question correctly,
and all in the same way. You therefore need to use simple language and avoid jargon that
may either not be understood, or be understood differently, by some groups or individuals.

4. Word your questions positively

This doesn’t mean that you have to be relentlessly enthusiastic. It just means that you
should ensure that your questions are worded so that if people feel good about something,
they ‘agree’ with the sentence or question.

This is for two reasons:

People may miss the negative, especially if most other questions are positively worded;
and
If you’re using a Likert-type scale (which answers on a scale from ‘Strongly agree’ to
‘Strongly disagree’), people may get confused about which way to reply.
Some surveys deliberately shift backwards and forwards between negative and positive
statements, often about the same idea or subject. This is one way to test whether your
respondents are paying attention , or just automatically ticking or clicking on ‘Agree’, but
you do need to use it carefully.

For example, you might use phrases like ‘I always…’ and ‘I never…’ rather than ‘I do…’ and
‘I do not…’ because the difference is more obvious.

5. Avoid leading questions

A leading question is one which points the respondent towards a ‘right’ answer. Beware of
imposing your views on others, however inadvertently.
The Golden Rule

In designing your questions, you need to keep all these five principles in mind,
and then think:

“What do I want to do with the answer?”

You need to be confident that the answers that you get will enable you to do the
analysis that you want, and then take the action that you need as a result.

A Worked Example
You are designing a survey for the Parent-School Association
to find out how parents feel about their programme of social
events.

They want to know how they can improve their events and
whether people will support particular events, or would prefer new
ones, so that they can plan their social timetable for the next year.

The original question was:

In the last year, have you attended any PSA events?

Yes, 5 or more

Yes, 3-4

Yes, 1-2

No, they are at inconvenient times or did not appeal

No, I’m a new parent

No, other reason (please specify)

Which events that you attended would you attend again? (Tick
from the list of possible events).

This question is positively phrased, but does not give sufficient

information for planning events.

The answers will tell you which events people liked, but not which
they disliked, or what aspects they disliked, to support
improvement. It is particularly difficult if someone ticks ‘Yes, 5 or
 more’ in answer to the first question, and then only one event in
answer to the second.

On balance, despite the desire to avoid negative questions, it

might be better to add a question about which events, if any,
would not be supported in future, with an optional box for
reasons. That way, you will get information that will enable the
group to change their events to meet parents’ needs.

Types of Answer Scale

The answers to your questions may be two types: category
scales and continuous scales.

Category scales use only a few distinctions, and may be either ordered or unordered.

Ordered scales, as the name suggests, have a numerical relationship between the
possible answers, so that shuffling the scale makes a difference. An example is the UK
government’s ABC1 categorisation of socio-economic status, and the Likert scales
mentioned on our page on Quantitative and Qualitative Data.
Unordered, or nominal, scales include things like ethnic origin.
Continuous scales have a large number of possible responses, for example, a
temperature scale.

Piloting and Testing

Finally, once you have developed your survey or questionnaire, it is worth taking a
bit of time to test it.

Suitable tests come in several forms. You might, for example, share your survey with a
focus group (see our page on Focus Groups for more) of experts and obtain their views on
it.

Alternatively, you might carry out a small-scale pilot study. Here, you would distribute your
questionnaire to perhaps 20 or 25 people with known characteristics, usually part of your
study population, and test whether their replies show good reliability and validity.
You may, for example, get a large number of non-responses to one or more questions,
which suggests that those questions were hard to understand and might need to be
revised. Equally, if you are using a Likert scale, and you get a lot of ‘neither agree nor
disagree’ responses, you may be asking about issues that are of no importance to your
target population. Again, you may want to revise your questions to address other issues
instead.
Observational Research and Secondary Data
Part of: Research Methods

This page covers two major potential sources of research data: observational
research, and use of someone else’s already-published data, known
as secondary data.

For other sources of data collection see our pages: Sampling and Sample
Design, Surveys and Survey Design and Quantitative and Qualitative Research
Methods.

Observational methods of collecting data have been used by scientists and researchers
for many years. As far back as the Middle Ages, scientists were observing what happened
as a result of their experiments. Similar methods are widely used in all types of research,
from laboratory work right through to management research, and even fieldwork in the
jungle.
Secondary data may be used in both quantitative and qualitative research, but involves the
use of previously-published information for analysis. Such data may include historical
archives, company records, and census data.

Collecting Observational Data

Observations range from counting something that happens, for example,
‘pings’ on a computer screen, through to observing and/or coding behaviour.

Observations may be used in a laboratory or in the field, for example, in a meeting at

an office. They allow researchers to uncover things that are not known, or not
spoken about, which would not be disclosed by interviewing people or using
surveys. Examples might include the informal power relationships within a group.

Endless Observations

It will almost certainly be obvious that there are as many types of observational
data as there are phenomena in the world; probably more, because behaviour
 counts as many different types of data. It is therefore impossible to discuss
every type.

Methods of collecting are also wide-ranging.

Data tend to be either seen and written down, or recorded on a computer. Observations
may be made and recorded immediately as observations, or the data may be recorded ‘raw’
and analysed later. This technique is often used in management research to record
meetings for later analysis of the conversation.
There are also two different options for the observer: he or she may be either an observer
from the outside, or they may be a participant.

This status will change the experiment, often significantly.

For example, it is probably obvious that whether someone observing a meeting is an ‘outsider’ or
a member of the team will alter the behaviour of the team. People tend to be wary of ‘washing
dirty linen in public’, but may also say more to an outsider because they do not fear that it will
alter their position in the team.

Other Factors Affecting Observational Data

Data is Affected by Being Observed
In behavioural terms this is obvious: people behave differently when they’re being
observed.

For this reason, observations about behaviour tend to be stronger when a group is
observed constantly over a long period. Under these circumstances, the group gets used to
being observed, and starts to behave more naturally. However, this type of design lends
itself more to some types of research than others.
For example, it’s quite hard to design management research that allows you to sit in an
office observing how a team behaves all day, every day for several months. However,
researchers like Dian Fossey and Jane Goodall used this technique to observe wild apes.
At first, the animals were very wary and spent most of their time watching the researcher
but, as the animals became accustomed to the presence of a human, they started to ignore
them and behave naturally.
The same principle applies to ‘pure’ scientific research. For example, the very act of
observing an electron affects its location, which means that measurement alters the
experiment. You always need to be aware of ‘observer effects’.

Observational Data is Affected by what is Sampled

Even if you have managed to design research which allows you to watch everything
that goes on for several months, you are still going to select what you notice,
whether consciously or unconsciously. It’s human nature. The key is to notice all the
data that is relevant to what you are studying and not just the data that fits with your
hypothesis.

This is where recording and re-examining later is helpful as you can go over the data
several times and ensure that you have included everything relevant. Another way to avoid
this so-called observer bias is to involve someone else in your re-examination and coding.
Much management research involves two observers and four or more coders working in
duplicate on transcripts or recordings later as a way of resolving observer bias without
introducing inconsistencies of coding.

Alternatively, you can sample at various time intervals, such as every ten minutes or every
hour.

Using Secondary Data

In subjects such as history and classical studies, secondary data is usually the
only available source of information.

Data may include eye witness accounts, contemporary reports of events, or

later reports. Historians generally give the greatest credence to the first, then
the second, and finally the third, although there is a place for all of them in
research.

For example, a formal record of an event, created for an official purpose, may not be
directly contemporaneous, but may draw on all available eye witness accounts, and
therefore give a better picture of events than any single eye witness account. In
general, government and formal company documents are higher quality than
personal documents, but you always need to be aware of why they were written.

In more scientific research, secondary data is often regarded as ‘second-best’ although it

is widely used particularly for public health and epidemiological research. Suitable data
sources for such research include national and international health surveys, often funded by
governments. The quality of such data depends on:
The size of the sample: the larger, the better, because the answer will be more precise
(and see our page on Sampling and Sample Design for more); and
The quality of the data collection, including how representative the sample is of the
population as a whole but also whether any bias has crept in during data collection.
 As a general rule of thumb, you can place high reliance on a large-scale survey carried out by a
highly reputable research institute and funded by government. Smaller-scale studies are less
reliable.

In social sciences, including management and business research, the position on

secondary data is more nuanced. Many studies will draw on some kind of secondary data,
but often supplement with primary data as well.

Examples of secondary data in these fields include:

Financial databases, such as archived company accounts;

Collections of newspaper reports; and
Census data.
It is important to assess the quality of the information before use, which depends on
several factors including its completeness and accuracy, and what information was
collected.

All these will depend on the purpose for which the information was collected in the
first place. Generally, if the purpose for which the data was collected is similar to the
purpose of your research, you are likely to find the data useful and be able to rely on it in
your study.

You also need to be aware of any changes to the data series over time, for example,
when one particular item was redefined to fit a different purpose. These may affect what
time period you can study, or make one time period less comparable with another.

In conclusion…
Observational research and secondary data both have a place to play in all
types of research.

As with any research design, the important aspect is to be guided by your

research questions to draw on the data that will answer those, and also to
assess the quality of your chosen methods to identify strengths and limitations.
Analysing Qualitative Data See also: An Introduction to Research Methods

Our pages on Quantitative and Qualitative Data and Collecting Qualitative

Data explain the various methods of collecting data. This page details how to
make sense of that data once collected.

There are many possible techniques to use, but what is important is that the
technique that you use is consistent with the philosophical view that underpins
your research.

Many analytical systems can be used for several different sorts of data, so the choice
of which to use is fairly subjective. It will depend on the philosophy, and also on your
own skills and preferences.

Systems for Analysis of Qualitative Data

Involving Language
Often, the output from qualitative research will be in the form of words.

For example, you may have collected data from or in written texts, or through in-depth
interviews or transcripts of meetings. According to Easterby-Smith, Thorpe and Jackson, in
their book Management Research, there are six main systems of analysis for language-
based data, which may also be used for other types of data.

1. Content Analysis
Here, you start with some ideas about hypotheses or themes that might emerge, and
look for them in the data that you have collected. You might, for example, use a colour-
coding or numbering system to identify text about the different themes, grouping together
ideas and gathering evidence about views on each theme.

2. Grounded Analysis
This is similar to content analysis, in that it uses similar techniques for coding. However,in
grounded analysis, you do not start from a defined point. Instead, you allow the data
to ‘speak for itself’, with themes emerging from the discussions and
conversations. In practice, this may be much harder to achieve because it requires you to
put aside what you have read and simply concentrate on the data.

Some people, such as Myers-Briggs 'P' types, may find this form of analysis much easier to
achieve than others. See our page on Myers-Briggs Type Indicators for more.

These first two approaches are not really as distinct as the description might
lead you to believe.

Instead, the pure approaches lie at opposite ends of a spectrum. For example, a
pure content analysis approach would have fixed themes. However, if more
information emerges from the data that does not fit with the pre-identified
themes, you may want to update and adapt your themes in the course of the
research. This approach is moving towards a hybrid approach, and perhaps a
more pragmatic approach than either pure system.

3. Social Network Analysis

This form of analysis examines the links between individuals as a way of
understanding what motivates behaviour.
It has been used, for example, as a way of understanding why some people are more
successful at work than others, and why some children were more likely to run away from
home. This type of analysis may be most useful in combination with other methods, for
example after some kind of content or grounded analysis to identify common themes about
relationships. It’s often helpful to use a visual approach to this kind of analysis to generate a
network diagram showing the relationships between members of a network.

4. Discourse Analysis
This approach not only analyses conversation, but also takes into account the social
context in which the conversation occurs, including previous conversations, power
relationships and the concept of individual identity. It may also include analysis of written
sources, such as emails or letters, and body language to give a rich source of data
surrounding the actual words used.

5. Narrative Analysis
This looks at the way in which stories are told within an organisation or society to try
to understand more about the way in which people think and are organised within
groups.

There are four main types of narrative:

 bureaucratic, which is highly structured and logical, and often about imposing control;
quest, where the ambition is to have the most compelling story and lead others to
success;
chaos, where the story is lived, rather than told; and
postmodern, which is rather like chaos narratives, in that it is lived, but where the
‘narrator’ is aware of the story and what they are trying to achieve.

6. Conversation Analysis
This is largely used in ethnographic research. It assumes that conversations are all
governed by rules and patterns which remain the same whoever is talking. It also
assumes that what is said can only be understood by looking at what went before and after.

Conversation analysis requires a detailed examination of the data, including exactly which
words are used, in what order, whether speakers overlap their speech, and where the
emphasis is placed. There are therefore detailed conventions used in transcribing for
conversation analysis.

Like content and grounded analysis, discourse, narrative and conversation

analysis can be considered as on a spectrum of systems for analysing forms of
language. Which you use will depend on what you want to achieve from the
analysis.

Computer-Aided Analysis
There are many computer packages designed to support and assist with the analysis
of qualitative (language-based) data, these include NVivo, Atlas.ti and the like. Their use is
beyond the scope of this page, but they are widely used to analyse large quantities of data,
reducing the pressure on a researcher to read and code everything him- or herself.
If you think that your research might need to use a package of this type, you are probably
best discussing it with your supervisor or a colleague who has experience of using the
package and can advise you about its use.

A Word of Warning

This page is necessarily only a brief summary of the techniques that can be
used to analyse language-based qualitative data. It is likely to be sufficient to
give you an idea of whether the technique will be useful.

However, if you decide to use any of the techniques or systems mentioned here,
you should read more about the technique in question, and discuss your plans
Simple Statistical Analysis See also: Designing Research

Once you have collected quantitative data, you will have a lot of numbers. It’s
now time to carry out some statistical analysis to make sense of, and draw
some inferences from, your data.

There is a wide range of possible techniques that you can use.

This page provides a brief summary of some of the most common techniques
for summarising your data, and explains when you would use each one.

Summarising Data: Grouping and Visualising

The first thing to do with any data is to summarise it, which means to present it in a
way that best tells the story.

The starting point is usually to group the raw data into categories, and/or to visualise it. For
example, if you think you may be interested in differences by age, the first thing to do is
probably to group your data in age categories, perhaps ten- or five-year chunks.
One of the most common techniques used for summarising is using graphs, particularly bar
charts, which show every data point in order, or histograms, which are bar charts grouped
into broader categories.
An example is shown below, which uses three sets of data, grouped by four categories.
This might, for example, be men, women, and ‘no gender specified’, grouped by age
categories 20–29, 30–39, 40–49 and 50–59.
An alternative to a histogram is a line chart, which plots each data point and joins them up
with a line. The same data as in the bar chart are displayed in a line graph below.

It is not hard to draw a histogram or a line graph by hand, as you may remember from
school, but spreadsheets will draw one quickly and easily once you have input the data into
a table, saving you any trouble. They will even walk you through the process.

Visualise Your Data

The important thing about drawing a graph is that it gives you an immediate
‘picture’ of the data. This is important because it shows you straight away
whether your data are grouped together, spread about, tending towards high or
low values, or clustered around a central point. It will also show you whether you
have any ‘outliers’, that is, very high or very low data values, which you may
want to exclude from the analysis, or at least revisit to check that they are
correct.

It is always worth drawing a graph before you start any further analysis, just to
have a look at your data.
You can also display grouped data in a pie chart, such as this one.

Pie charts are best used when you are interested in the relative size of each group, and
what proportion of the total fits into each category, as they illustrate very clearly which
groups are bigger.

See our page: Charts and Graphs for more information on different types of graphs and charts.

Measures of Location: Averages

The average gives you information about the size of the effect of whatever you are testing,
in other words, whether it is large or small. There are three measures of average: mean,
median and mode.

See our page on Averages for more about calculating each one, and for a quick calculator.

When most people say average, they are talking about the mean. It has the advantage that
it uses all the data values obtained and can be used for further statistical analysis.
However, it can be skewed by ‘outliers’, values which are atypically large or small.

As a result, researchers sometimes use the median instead. This is the mid-point of all the
data. The median is not skewed by extreme values, but it is harder to use for further
statistical analysis.

The mode is the most common value in a data set. It cannot be used for further statistical
analysis.
The values of mean, median and mode are not the same, which is why it is really important
to be clear which ‘average’ you are talking about.

Measures of Spread: Range, Variance and Standard Deviation

Researchers often want to look at the spread of the data, that is, how widely the data are
spread across the whole possible measurement scale.
There are three measures which are often used for this:

The range is the difference between the largest and smallest values. Researchers often
quote the interquartile range, which is the range of the middle half of the data, from 25%,
the lower quartile, up to 75%, the upper quartile, of the values (the median is the 50%
value). To find the quartiles, use the same procedure as for the median, but take the
quarter- and three-quarter-point instead of the mid-point.

The standard deviation measures the average spread around the mean, and therefore
gives a sense of the ‘typical’ distance from the mean.

The variance is the square of the standard deviation. They are calculated by:

1. calculating the difference of each value from the mean;

2. squaring each one (to eliminate any difference between those above and below the
mean);
3. summing the squared differences;
4. dividing by the number of items minus one.
This gives the variance.
To calculate the standard deviation, take the square root of the variance.

Skew
The skew measures how symmetrical the data set is, or whether it has more high values,
or more low values. A sample with more low values is described as negatively skewed and
a sample with more high values as positively skewed.

Generally speaking, the more skewed the sample, the less the mean, median and mode will
coincide.

More Advanced Analysis

Once you have calculated some basic values of location, such as mean or
median, spread, such as range and variance, and established the level of skew, you can
move to more advanced statistical analysis, and start to look for patterns in the data.
Averages: Mean, Median and Mode See also: Percentages

The term ‘average’ occurs frequently in all sorts of everyday contexts – you
may say ‘I’m having an average day today’ meaning your day is neither
particularly good nor bad, it is about normal. Similarly we may refer to people,
objects and other things as ‘average’.

The term 'average' refers to the ‘middle’ or ‘central’ point; when used in
mathematics the term average refers to a number that is a typical
representation of a group of numbers (or data set). Averages can be
calculated in different ways - this page covers the Mean, Median and Mode.
We include an averages calculator and explanation and examples of each
type of average.

The most widely used method of calculating an average is the ‘mean’. Most of the
time when the term ‘average’ is used in a mathematical sense it refers to
the meanaverage.

Quick Guide:

To calculate the Mean

Add the numbers together and divide by the number of numbers.
(The sum of values divided by the number of values).

To determine the Median

Arrange the numbers in order, find the middle number.
(The middle value when the values are ranked).

To determine the Mode

 Count how many times each value occurs the highest is the mode.
(The most frequently occurring value)

Mean, Median and Mode Calculator

Use this calculator to work out the mean, median and mode of a set of numbers.

Mean
The mathematical symbol or notation for
mean is ‘x-bar’. This symbol appears on
scientific calculators and in mathematical
and statistical notations.

The ‘mean’ or ‘arithmetic mean’ is the most commonly used form of average. In order to
calculate the mean average a set of related numbers (or data set) is required. At least two
numbers are needed in order to calculate the mean.

The numbers need to be linked or related to each other in some way to have any
meaningful result – for instance, temperature readings, the price of coffee, the number of
days in a month, the number of heartbeats per minute, student's test grades etc.

To find the (mean) average price of a loaf of bread in the supermarket, for
example, first record the price of each type of loaf:

White: £1
Wholemeal: £1.20
Baguette: £1.10
Next, add (+) the prices together £1 + £1.20 + £1.10 = £3.30

Then divide (÷) your answer by the number of loaves (3).

£3.30 ÷ 3 = £1.10.

The average price of a loaf of bread in our example is £1.10.

Mode
The Mode is the most frequently occurring value in a set of values. The mode is
interesting as it can be used for any type of data, not just numbers.

In this example, assume that you have bought a pack of 100 balloons, the pack is made up
of 5 different colours, you count each colour and find that you have:

18 Red
12 Blue
24 Orange
25 Purple
21 Green

The mode of our sample of balloons is purple as there are more purple balloons (25)
than any other colour balloon.

To find the mode of the number of days in each month:

Month Days

January 31
February 28
March 31

April 30
May 31

June 30
July 31
August 31

September 30
October 31

November 30
December 31

7 months have a 31 days, 4 months have a total of 30 days and only 1 month has a total of
28 days (29 in a leap year).

The mode is therefore, 31.

Some data sets may have more than one Mode:

1,3,3,4,4,5 – for example, has two most frequently occurring numbers (3 & 4) this is known
as a bimodal set. Data sets with more than two modes are referred to as multi-modal data
sets.

If a data set contains only unique numbers then calculating the mode is more
problematic.

It is usually perfectly acceptable to say there is no mode, but if a mode has to be found
then the usual way is to create number ranges and then count the one with the most points
in it. For example from a set of data showing the speed of passing cars we see that out of
10 cars the recorded speeds are:

40, 34, 42, 38, 41, 50, 48, 49, 33, 47

These numbers are all unique (each only occurs once), there is no mode. In order to find a
mode we build categories on an even scale:

30--32 | 33--35 | 36--38 | 39--41 | 42--44 | 45--47 | 48--50

Then work out how many of the values fall into each category, how many times a number
between 30 and 32 occurs, etc.
30--32 = 0
33--35 = 2
36--38 = 1
39--41 = 2
42--44 = 1
45--47 = 1
48--50 = 3

The category with the most values is 48--50 with 3 values.

We can take the mid value of the category to estimate the mode at 49.
This method of calculating the mode is not ideal as, depending on the categories you
define, the mode may be different.
Graphs and Charts See also: Averages

A picture, so they say, will tell a thousand words. But what about a graph or chart?

A good graph or chart can show as much as several paragraphs of words. But how do you
choose which style of graph to use?

This page sets out some of the basics behind drawing and creating good graphs and charts. By
‘good’, we mean ones that show what you want them to show, and don’t mislead the reader.

Types of Charts
There are several different types of charts and graphs. The four most common are probably line
graphs, bar graphs and histograms, pie charts, and Cartesian graphs. They are generally used
for, and best for, quite different things.

You would use:

Bar graphs to show numbers that are independent of each other. Example data might include things
like the number of people who preferred each of Chinese takeaways, Indian takeaways and fish and
chips.
Pie charts to show you how a whole is divided into different parts. You might, for example, want to
show how a budget had been spent on different items in a particular year.
Line graphs show you how numbers have changed over time. They are used when you have data
that are connected, and to show trends, for example, average night time temperature in each month of
the year.
Cartesian graphs have numbers on both axes, which therefore allow you to show how changes in
one thing affect another. These are widely used in mathematics, and particularly in Algebra.

Axes
Graphs have two axes, the lines that run across the bottom and up the side. The line along the bottom is
called the horizontal or x-axis, and the line up the side is called the vertical or y-axis.

The x-axis may contain categories or numbers. You read it from the bottom left of the graph.
The y-axis usually contains numbers, again starting from the bottom left of the graph.
The numbers on the y-axis generally, but not always, start at 0 in the bottom left of the graph, and move
upwards. Usually the axes of a graph are labelled to indicate the type of data they show.

Beware of graphs where the y-axis doesn't start at 0, as they may be trying to fool you about the data
shown (and there is more about this on our page, Everyday Mathematics).
Line graphs depict a point value for each category, which are joined in a line. We can use the data from
the pie chart as a line graph too.

You can see even more obviously that sales have fallen rapidly over the year, although the slow-down is
levelling out at the end of the year. Line graphs are particularly useful for identifying the point in time at
which a certain level of sales, revenue (or whatever the y value represents) was reached.
In the example above, suppose we want to know during which quarter sales first fell below 5. We can
draw a line across from 5 on the y-axis (red line on the example), and see that it was during quarter 2.

Cartesian Graphs
Cartesian graphs are what mathematicians really mean when they talk about graphs. They compare two
sets of numbers, one of which is plotted on the x-axis and one on the y-axis. The numbers can be written
as Cartesian coordinates, which look like (x,y), where x is the number read from the x-axis, and y the
number from the y-axis.

Warning!

Cartesian graphs do not always start at 0; quite often (0,0) is the mid-point of the graph.

Cartesian Graph - Worked Example

John is two years older than Mary, and their ages

added together equal 12. What age are they both
now?

We can solve this by drawing two lines, one of John’s age

compared with Mary’s, and one of the ages that add together to
Statistical Analysis: Identifying Patterns
See also: Simple Statistical Analysis

More advanced statistical analysis aims to identify patterns in data, for

example, whether there is a link between two variables, or whether certain
groups are more likely to show certain attributes.

This is in order to draw lessons from the sample that can be generalised to the
wider population.

For more on making sure that your sample is large enough to allow you to
generalise, see our page on Samples and Sample Design.

Relationships vs Differences

Research hypotheses can be expressed in terms of differences between

groups, or relationships between variables. However, these are two sides of the
same coin: almost any hypothesis could be set out in either way.
For example:
There is a relationship between gender and liking ice cream OR

Men are more likely to like ice cream than women.

Comparing Groups
Your first step is to identify your two or more groups. This will obviously depend on
your research question or hypothesis.

So if your hypothesis was that men are more likely to like ice cream than women, your two
groups are men and women, and your data is likely to be something like self-expressed
liking for ice cream on a scale of 1 to 5, or perhaps the number of times that ice creams are
consumed each week in the summer months.
You then need to produce summary data for each group, usually mean and standard
deviation. These may or may not look quite similar.

In order to decide whether there is a genuine difference between the two groups, you have
to use a reference distribution against which to measure the values from the two groups.
The most common source of reference distributions is a standard distribution such as the
normal distribution or t- distribution. These two are the same except that the standard
deviation of the t-distribution is estimated from the sample, and that of the normal
distribution is known.
You then compare the summary data from the two groups and decide the probability of
achieving that difference by chance. The lower the probability, the more likely it is that your
result is correct. This is referred to as statistical significance.

Types of Error

There are four possible outcomes from statistical testing:

The groups are different, and you conclude that they are different (correct
result)
The groups are different, but you conclude that they are not (Type II error)
The groups are the same, but you conclude that they are different (Type I
error)
The groups are the same, and you conclude that they are the same (correct
result).
Type I errors are generally considered more important than Type II, because they
have the potential to change the status quo.

For example, if you wrongly conclude that a new medical treatment is effective,
doctors are likely to move to providing that treatment. Patients may receive the
treatment instead of an alternative that could have fewer side effects, and
pharmaceutical companies may stop looking for an alternative treatment.

Choosing the Right Test

The test that you use to compare your groups will depend on how many
groups you have, the type of data that you have collected, and also how good
it is. In general, different tests are used for comparing two groups, and for
comparing three or more.

Our page Surveys and Survey Design explains that there are two types of answer
scale, continuous and categorical. Age, for example, is a continuous scale, although
it can also be grouped into categories.

Gender is a category scale.

For a continuous scale, you can use the means of the two groups that you are
comparing.
For a category scale, you need to use the medians.

Warning!

If you are not very confident about the quality of the data collected, for example
because the inputting was done quickly and cheaply, or because the data have
not been checked, then you may prefer to use the median even if the data are
continuous to avoid any problems with outliers. This makes the tests more
robust, and you can rely on the results more.

What Test?
Test Reference
Purpose Data Scale Average Test
Statistic Distribution
Continuous Mean t-test t t
Compare two All
groups Mann-Whitney U U
Category Median combination
test statistic
of ranks
Analysis of
Continuous Mean Variance F-ratio F
Compare three or (ANOVA)
more groups All
Kruskal-Wallis W
Category Median combination
Test statistic
of ranks
Source: Easterby-Smith, Thorpe and Jackson, Management Research 4th Edition

One- or Two-Tailed Test

The other thing that you have to decide is whether you are confident of the direction of the
distance. In practice, this boils down to whether your research hypothesis is expressed as
‘x is likely to be more than y’, or ‘x is likely to be different from y’. If you are confident of the
direction of the distance, then your test will be one-tailed. If not, it will be two-tailed.

Calculating the Test Statistic

For each type of test, there is a standard formula for the test statistic. For example, for the t-
test, it is:

(M1-M2)/SE(diff)

M1 is the mean of the first group

M2 is the mean of the second group

SE (diff) is the standard error of the difference, which is calculated from the standard
deviation and the sample size of each group.

The final part of the test is to compare the test statistic to that required to meet the desired
level of significance (usually 5% or 1%). This value is available from published statistical
tables. If the test statistic is that value or more, then the difference between groups is said
to be statistically significant at the 5% or 1% level.

NOTE: the significance level is sometimes called the p value, and expressed as p < 0.05 or p<
0.01.

Comparing Variables
Sometimes, you may want to know if there is a link between two variables. If so, you can
predict someone’s response to one variable by their response to the other.

A positive association means that high scores for one variable tend to occur with
high scores for the other.
A negative association means that high scores for one variable tend to occur with
low scores for the other.
There is no association when the score for one variable does not predict the score for
the other.
Such associations are also called correlations.

Seeing an Association

One of the best ways of checking for an association is to draw a line graph of
the data with the two variables on the x and y axes. Broadly speaking, if there is
an association, you will see it from the graph.
 Drawing a graph will also help you identify if there is a peculiar relationship,
such as a positive association for part of the data and a negative for the rest, as
shown below. This will show in a test as no correlation, but there is clearly some
sort of a relationship in this case.

Statistical Tests for Associations

Again, there are specific tests depending on whether you are using continuous,
categorical or ranked data.

For categorical data, you use the chi-squared test (also written χ2)
For continuous data it is the Pearson product-moment correlation
For ranks, use the Kendall rank order correlation.
Again, you need to work out the test statistic, and compare that with the value needed to
obtain the desired level of significance.

Warning! The Difference Between Correlation and Causation

A correlation is an association between two variables. It does not necessarily

imply that one causes the other. Both could be caused by something completely
different, or it could simply be that people who show one characteristic often
show the other.

For example, it could be that people who shop for groceries online buy more
ready-made meals than those who shop in store. However, it is unlikely that the
act of buying online causes the purchase of more ready-meals. It is more likely
that those who shop online are short of time, and so buy more convenience
food, or possibly simply that younger people are both more likely to shop online
and more likely to buy convenience food.
Multivariate Analysis See also: Research Methods

Our pages Simple Statistical Analysis and Identifying Patterns in Dataexplain

some of the simpler techniques used for statistical analysis. This page
discusses some of the more advanced techniques, involving several variables
and not just one or two.

In real life, as opposed to laboratory research, you are likely to find that your
data are affected by many things other than the variable that you wish to test.
There are correlations between items that you’ve never considered, and the
world is complex.

The purpose of advanced statistical analysis is to simplify some of the

relationships, while making a more effective model of what you are seeing.

There are Four Ways to Simplify Analysis

Design
Using Sub-Samples
Using Statistical Controls
Multivariate Analysis

1. Design
You can design your research so that causal factors are made independent of each
other. For example, if you think that there may be a link between age and salary, then a
random sample of employees will risk combining the effects of both. If, however, you divide
the population into groups by age, and then randomly sample equal numbers from each
group, you have made age and salary independent.

2. Using Sub-Samples
Here, you select your sample to be equal on any potentially confounding factors.For
example, job type may affect pay, so if you want to study the effects of another factor on
pay, you could select only people doing the same job.

3. Using Statistical Controls

If you suspect that three variables may be linked, you can control for one to test for
correlations between the other two. Effectively, you adjust the statistical value of the
control to be constant, and test whether there is still a relationship between the other two
variables. You may find that the observed relationship remains high (it is real), or reduces
considerably (there is probably no real relationship). There is a third case: where there is no
relationship until you control the third variable, which means that the control variable is
masking the relationship between the other two.

4. Multivariate Analysis
Multivariate Analysis includes many statistical methods that are designed to allow
you to include multiple variables and examine the contribution of each.

The factors that you include in your multivariate analysis will still depend on
what you want to study. Some studies will want to look at the contribution of
certain factors, and other studies to control for those factors as (more or less) a
nuisance.

Two types of variables

In multivariate analysis, the first thing to decide is the role of the variables.

There are two possibilities:

The variable causes an effect: predictor variable

The variable is affected: dependent variable
This is a function of your model, not of the variables themselves, and the same variable
may be either in different studies.

The relationships between variables are usually represented by a picture with

arrows:
You can also observe variables directly, or infer them from what is happening these are
known as latent variables.

Example: Success at School

It is hard to measure ‘success at school’: it is a latent variable.

You might decide that 'success at school' consists of academic success,

together with some measure of social success (perhaps average duration of
friendships, or size of ‘friendship group’) plus one of effort put in. These are your
observed variables.

The measurement model examines the relationship between the observed and
latent variables.

The idea behind such models is that there are correlations between the observed
variables.

These correlations are assumed to be caused by common factors. The greater the
influence of the common factors (the factor loading), the higher the correlations between
the variables.
Cronbach’s alpha is used to measure the correlations between variables. A value of 0.70
or more gives a good level of reliability to the model.

Methods of Analysis
There are a variety of methods of analysis for measurement models like this.
They include Confirmatory Factor Analysis and Exploratory Factor
Analysis, and are usually carried out by computer.

The details of how to carry out each one are beyond the scope of this page,
but the basic idea is that they measure how much of the variation seen in the
overall construct is caused by each factor.

Causal Models
Causal models look at the way in which variables relate to each other. While it is not
possible to prove causality beyond doubt, causal models allow you to say whether
the suggested relationship fits the data, and how well.

The strength or weakness of any causal model is the selection of the variables.
If you miss out a major causal factor, then your conclusions will be either limited
or incorrect. It is therefore worth taking time on defining your model as carefully
as possible.

There is a balance to be struck between simplicity and including more variables

to obtain a better fit. Obviously you do not want to miss out a major causal
variable, and including more variables will always give a better fit. But you need
to consider whether the additional complexity is worth it for the gain in quality of
the model.

Suitable analysis methods for causal models tend to be what is called generalised
linear models, which include logistic regression analysis, multiple regression
analysis, multivariate analysis of covariance (MANCOVA) and multivariate analysis of
variance (MANOVA).

All these methods give you a measure of how much of the variation in the dependent
variables is caused by the predictors, and thus whether your model is any good.

Again, there are computer packages, such as SPSS, which can do these analyses for you,
but do make sure that you understand what you’re doing and are interpreting the results
correctly.

Structural Equation Modelling brings together measurement models and causal models.
It is a computer-modelling technique that fits a structural equation to the model. This
technique is pretty complicated, but in essence compares possible models and identifies
the one that best fits the data.

A Complex Area
The world is a complex place, and sometimes the only way to understand what’s
going on is to use advanced statistical techniques for modelling.

However, these too are complex and you should not embark on them without understanding
the basics. If you don’t, then it’s a good idea to consult someone who does, usually a
statistician. Even if you’ve used the technique before, it’s still a good idea to get a