11
10
Analyzing qualitative data
lntroduction lOO
Coding and categorizing 100
Anolysis of conversotions, discourse ond documents 104
Chapter objectives
After reading this chapter, you should
• hove a first orientotion about the mojor approoches in onolysing
quolitotive data;
• see the specific issues concerning reseorch designs in using these
methods; and
• hove some orientotion obout when to use which methods of
analysis in qualitative research.
lntroduction
The methods briefly presented in the preceding chapters are focused on collecting
data (like interviews or focus groups). Or they consist of a more integrated
approach (a link between a specific sort of data anda way of analyzing it- visual
data or discourse analysis, for example). In this penultimate chapter, we will tum to
approaches in which the analysis of qualitative data is in the foreground. Here we
will find two different ways - coding and categorizing, which can be applied to
qualitative data in general, whereas later a more specified analysis of data will be
presented. At the end of this chapter we will try to bring the different issue of this
book - the design issues and the methods - together for sorne kind of outlook.
Coding and categorizing
Coding and categorizing are ways of analyzing that can be applied to all sorts of
data and are not focused on a specific method of data collection. This is not the
(
/
(
\
(
(
(
(
(
(
(
(
e
(
'
(
(
Analyzing qualitative data
only way of analyzing data, but it is the most prominent one, if the data result
from interviews, focus groups or observations. In particular, if computers are
used for analyzing qualitative data, one or the other form of coding will be
applied. The main activities are to search for relevant parts of the data and to analyze
them by comparing them with other data and by naming and classifying
them. Through this process, a structilre in the data is developed as a step towards
a comprehensive understanding of the issue, the field, and last but not least the
data themselves. It is suggested to interlink data collection with their analysis in
order to profit from insights from the analysis for the process of collecting
further data.
Research perspective and theory
The background of coding and categorizing can be realist or constructivist
(Gibbs, 2007). The aim is often to develop a theory and therefore the categories
for coding the material are developed from the material rather than from existing
theories, although this way is possible and usual as well. The approaches need
materials at hand - existing documents, transcriptions of interviews, focus
groups or interactions. They can also be applied to visual material or Internet
material.
Research questions
These approaches are open to all sorts of research topics. They are less appropriate
where formal structures (like organizations of specific forms of talk as in conversation
analysis) are the main issue. Although narratives can be analyzed with coding
and categories as well (see Gibbs, 2007, chap. 5), an analysis interested in the interna!
structure and the gestalt of the narrative is more difficult with coding.
Sampling
When this form of analysis starts, sampling of cases and materials is often
(
'-. already completed. However, sampling in qualitative research can and maybe
even should be based on the progress of the analysis of the data collected so far.
"- Therefore, this form of analysis can have an impact on the sampling of cases and
materials. During the analysis, sampling in the case and in the material is an
(
\_
l
l
l
important element.
Comparison
Here we can plan comparisons on three levels: Within a category- what do we
find in different interviews, for example, as relevant for a specific category? 101
102
Designing qualitative research
Within a case - what does the interviewee say about different issues, how
consistent or contradictory are statements across several categories? Between
cases - how different or similar are the responses of various interviewees on the
level of one topic/category or on the level of the whole interview? Comparisons
on the basis of focus groups or observations can be made on similar levels.
Comparison here can be facilitated by ordering and structuring - like building
a hierarchy.between categories, by developing tables for displaying several
cases and categories allowing case-by-case or chronological comparisons (see
Gibbs, 2007).
Generalization
Constant comparison of materials is an important step on the way from the
single caseto more (or less) general statements drawn from analyzing the data.
To avoid over-generalization, researchers should carefully reflect the boundaries
of their data and the sample of people (or materials) they are based on.
Triangulation
Coding and categorizing can be combined with quantitative analyses of standardized
data. It may also refer to different sorts and sources of qualitative data.
Also, Gibbs (2007) argues for ｾｲｩ｡ｮｧｵｬ｡ｴｩｮｧ＠ the perspectives of different
researchers on the data as well as seeking for respondent validation - to integrate
participants' perspectives on the data into the final analysis.
Quality
Central for the quality of data analysis, according to Gibbs (2007), is that the
researchers are reflexive in their practice with the data by critically assessing their
own roles as researchers as well as the data and the findings and conclusions
drawn from them. Reliability can be increased by rechecking the transcripts and
by cross-checking the codes.
(
(
(
(
(
(
(
(
(
(
ｾｍｧ＠ (
In this context, writing is not only referring to presenting results in a report to
audiences. All sorts of research materials like memos, protocols, field notes and (_
research diaries can become relevant for the analysis and are products of the
researchers' writing processes. In reports about the research, a good relation
between categories that are presented, analysis and conclusions based on them,
and excerpts from the 'raw' material is an important goal.
(
\.,
(
\.
Analyzing qualitative data
Basic design
In most cáses, coding and categorizing will be based on a comparative design. It
can be applied in retrospective or longitudinal design as well as in snapshots,
because this is decided more by the sort of data (and their focus and collection)
that is used for the study.
Resources and stepping-sfones
Analyzing qualitative data in most cases is a time-consuming step in the research,
which needs most of the resources of a project. A good transcription will make
this step easier to handle, but you will need to spend time or money on this step,
if it is to be a good basis for interpretation. Using computers can be helpful for
supporting qualitative analysis in two cases in particular: if you are already experienced
with the software you want to use and if you have a bigger corpus of data.
A major problem to avoid is to have an unclear research question and not to have
focused it again when you begin your analysis. Another problem may arise when
you find out that your data actually do not include those statements or issues you
discover as important for your analysis in later stages.
Ethics
In the analysis of qualitative data, anonymity and confidentiality are central issues
from the angle of ethics - in transcription, in analysis itself, and most of all in presenting
results and excerpts from the data. For example, researchers should be sure
that those who do the transcription keep up standards of confidentiality. If you
feed your results back to participants, you should take care that they do not feel
embarrassed or hurt by your interpretation, without giving them a chance to
reflect with you on what was disturbing thern. Interpretations should be fair to the
research participants (and rnaybe the institutions) in the research.
Example
In our project on homeless adolescents, we used thematic coding (Flick, 2006) for
·, data analysis, beginning with all statements referring toan area (e.g. meaning of
health) for every interviewee. Comparative dirnensions are defined across cases
for finding commonalities and differences between the various interviews. Cases
are grouped along these dimensions and analyzed for specific combinations of
features. Contrasting cases allow comparing the cases in one group for similarities
and comparing cases across the groups for existing differences among
them. Interpretive and practice pattems can be found and analyzed in this
way. Stepping-stones are missing statements by one or more participants about
(
l...
specific issues.
103
104
Designing qualitative research
Conclusion
Using codes and categories for analyzing qualitative data is quite flexible in the
sort of data to be used and how the study is designed. It is mostly planned with
the aim of comparing data and cases. Triangulation should include different
researchers working on the same material for extending the perspectives.
Presentations should take the relation between original material and categories or
dimensions into account so that the analysis becomes transparent.
Analysis of conversations, discourse
and documents
In these forms of qualitative analysis, data collection is often limited to selecting
material (like documents or newspaper articles) and to documenting everyday
routines (like tape-recording a conversation between doctor and patient). Here,
research interests often focus more on formal aspects - how is a conversation
started, continued and ended, what are the structural aspects of a document like a
patient record - than on content. In discourse analysis, the stress is more on content,
however, compared to conversation analysis. In discourse analysis we also
find the use of interviews, observations and focus groups as sources for materials
to analyze.
Research perspective and theory
The theoretical background is in most cases ethnomethodology and discourse theory
and the interest is in how communication and practices are constructed
in everyday life in concrete circumstances. Therefore, the single actor is less
focused than interaction processes. On the epistemological level, these
approaches are based on social constructionism and theories of (written or spoken)
language use in practica! contexts.
Research questions
The primary research question of these kinds of research is: how is a specific
issue constructed in sorne sort of communication and which 'methods' do participants
in this communication use for this construction? Thus, documents are
always produced by someone and for sorne audience, for sorne purpose and by
using sorne communicative devices in order to create a certain format of information.
This can be a specific way of communicating with a patient about her illness
in a doctor-patient communication, which aims at having a diagnosis, a
treatment and a plan for proceeding further at the end of the session. This can
also be a record produced about this patient, her history as a patient, diagnosis
(
r·
'\
(
(
'
(
(
(
(
(
(
(
1
\..
(
{
'
(
"-
(
'Analyzing
qualitative data
treatments and prognosis or a public discussion about this specific illness in the
newspaper with the purpose of creating a moral discourse about the disease and
the people concemed with it. In each case the methods of producing the specific
document or result are in the focus of this sort of research.
Sampling
As Rapley (2007) holds, the strategy here is building an archive of materials for
analyzing the discourse or documents. Selection or sampling then refers to the
documents to choose for this archive. Sampling in this kind of research first aims
at constructing a corpus of material, which then can be a starting point for sampling
inside the materials in this corpus.
Comparison
In conversation analysis, comparison is in many cases oriented towards a more
general model (of how a conversation or a telephone call works), which is
juxtaposed to the concrete case that is being studied. Often you will set up an
inventory of several exemplars (like a number of beginnings of a counselling
conversation) and then compare them with each other to find regular and structural
principies in them. Comparison of documents is either focusing the contents
or the structure of a series of examples.
Generalization
Discourse analysis is more interested in case studies, so that we can mostly find
interna! generalization here, whereas conversation analysis in most cases is interested
in identifying general principies of talk and conversations. In this aim, a
generalization from case studies to comparing different cases to general models
is implied.
Triangulation
Rapley (2007) mentions the combination of analyzing documents and conversations
in this context. A different way is to combine conversation analysis with interviewing,
for example for analyzing professional practices by analyzing conversations and
professional knowledge by doing and analyzing interviews (see below).
Quality
In studying conversation, discourse and documents, Rapley (2007) holds
that there is no claim about truth in interpretations that can be justified. Rather,
105
106
Designing qualitative research
(
researchers should try to convince the reader that their interpretations are
plausible and credible by showing how they are based on materials and.their
analysis. Another approach to quality here is to validate documents in team meetings
of the research group. In general it should become visible in the results and
their presentations, how the researchers have checked and rechecked their conclusions
against the material and especially against cases and materials that are
deviant from their findings, conclusions and generalizations (as in a typology: see
Flick, 2007, for managing diversity in qualitative analysis). (
Writing
Depending on the research questions and the materials used in the study, different
forms of writing can be appropriate here. An important issue here is how
to demonstrate how conclusions are based in the material, and in particular in
(
(
conversation analysis with its sometimes very detailed transcriptions, how to (
present excerpts from the material in an accessible and readable way. Too detailed
a transcription can obstruct the perspective on the content and context of what is ( .
said; too simplified presentations of transcripts may undermine (the impression
of) exactness. In discourse analysis in particular, writing and analysis are very
closely linked to each other and writing up has an impact on the analysis (new
ideas, categories or clarifications).
Basic design
Most research in these areas is providing a snapshot- analyzing materials in the here
and now for what contents it includes or how it is structured. Conversation analysis
is mostly based on comparing different examples (of counselling interaction, for
example), whereas many discourse analyses are based on case studies. In document
analysis we often find a retrospective approach (for example, how has the documentation
of diagnoses of mental illness changed over the years?).
Resources and stepping-stones
In conversation analysis a major resource you need is time (and money) for
transcription of what has been recorded before (with very good equipment
if possible - another necessary resource). Finding and accessing the relevant
material for describing the discourse about an issue can be difficult and
time- consuming as well. This will be the case, in particular, if you want to
have (all) the relevant material and not just a single example or an awkward
selection. IIl. conversation and document analysis, a danger is to forget about
the contents and their meaning in analyzing basically the structure of the
material.
(
(
(
(
(
(
(
(
(
\.
(
•,
(
'-.
(
(
(
(
\
(
\,
/
'-
1
\.
(
\
f
Analyzing qualitative data
Ethics
If you base your research on (audio or video) recording for producing data, you
will have to make sure that the participants know about and have consented to
being recorded and that they have the chance to make you stop your recording if
necessary. With the material you have to take care of anonymity issues - that
nobody is identified in or through your research and that you do not talk openly
about your participants. Here again, such issues become more difficult once you
work with vulnerable people or sensitive topics (see Rapley, 2007, chap. 3).
Example
In a study in the field of community psychiatry, I could show, for counselling conversations,
how starting into conversation is organized so that a consultation begins
and not something different. In the conversations analyzed, such a start may be
designed rather open-ended (e.g. 'What made you cometo us?' or 'And what is it
about?' or 'What is your desire?'). In other cases, they name the (given) topic for
the counselling, or specific characteristics in the way the counselling conversation
carne into being. These openings, which begin the actual counselling relationship
and delimit it against other forms of talk, are sometimes linked to explanations
about the way the conversation carne about. These explanations are specific for the
situation (e:g. 'So, your brother gave me a call').
In analyses of the ending of first contacts·in counselling processes, I could
show two tasks to be achieved. A ｾｩｭ･ｬｹ＠ ending of the conversation has to be
ensured. At the same time, the counsellor has to guarantee the continuation of the
relation. This analysis could show which formal steps counselling conversations
ran through more or less regularly. It could also show how these steps not only
built up the conversation in itself, but also were influential in processing the
clients and their cases- regardless of the specific content of their problems. So,
the analysis was more formal than content-oriented, but shows the construction
of cases in the conversations. Sampling was oriented towards cases of conversation
and then to formal parts (beginnings, endings) in it; comparison of these parts
was the aim of the design and analysis. Ethical concems are about having the
informed consent of clients and the counsellors and to maintain the privacy of
ｾ＠ both. This conversation analysis was triangulated with interviewing the counsel-
"-
\.
lor, whose consultation was analyzed.
Conclusion
These analyses often focus formal parts of verbal exchange more than the content,
although discourse analysis is interested in both - form and content. Basic designs
are comparative snapshots in most cases, although documents are sometimes studied
with a retrospective focus. From an ethical point of view, anonymity and the
sometimes confrontational character of the analysis should be taken into account,
107
108
Designing qualitative research
in particular if you use the results (and examples) for training or other practica!
purposes.
Key points
• These approaches to analyzing qualitative data are either quite open
in the theoretical approach (as in coding) or come with a background
(ethnomethodology in conversation analysis).
• They stress either content or formal structures.
• In both cases, ethics demand a special care for the anonymity of the
participants in the analysis.
Furthe r reading
The fellowing book will go into more detail about the rnethod outlined in thi
chapter:
Gibbs, G. R. (2007) Analyzing Qualitative Data (Book 6 of The SAGE Qualitative
Research Kit). London: Sage.
Rapley, T. (2007) Doing Com er ation, Di cour e and Docurnent Anal is (Book 7
of The SAGE Qualitative Re earch Kir). London: Sage.
(
(
(
(
(
(
(
(
(
(
1
\.