Concept mapping as a methodical and transparent data analysis process 1 Abstract A single source of qualitative data collected using in- depth interviews in a grounded approach was used to compare the results of the more traditional matrix method of data analysis using NVivo, and the less common network method of concept mapping. Both forms of analysis produced similar results, and it is suggested that researchers consider using concept mapping as a valid method of qualitative analysis that, among other benefits, provides a clear audit trail for verification and collaboration. Keywords: qualitative; grounded; concept mapping; matrix analysis; NVivo Introduction In this chapter, the results of two different methods of analysis of the same set of qualitative data are directly compared. The research aim for this case study was to identify the dimensions of the innovation capability construct using a grounded approach. For this research, innovation capability was explored in a homogenous industry sector, (viz., small general hotels in Australia). The qualitative data used in this research was obtained from relatively unstructured in- depth interviews with 36 hotel owner/managers. Two researchers independently analysed interview transcripts using two V2-6 6Nov14
44
Embed
Idea Networkingideanetworking.com.au/docs/ideanetworking/Concept... · Web viewThe innovation capability (IC) literature describes this construct as the capacity of a firm to develop
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Concept mapping as a methodical and transparent data analysis process 1
Abstract
A single source of qualitative data collected using in-depth interviews in a grounded approach was used to compare the results of the more traditional matrix method of data analysis using NVivo, and the less common network method of concept mapping. Both forms of analysis produced similar results, and it is suggested that researchers consider using concept mapping as a valid method of qualitative analysis that, among other benefits, provides a clear audit trail for verification and collaboration.
The independently-conducted analyses yielded similar findings. This case study demonstrates
that concept mapping has certain advantages over the more traditional matrix analysis in that
it does not depend on the researcher a priori identifying nodes or categories, the visual output
allows a quick assessment of the results and the possibility to evaluate construct validity and
inter-dimensional relationships between clusters or coding categories, it provides a clear audit
trail, and takes less time.
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 2
The Research Question, the Innovation Capability Construct, and the Research
Methodology
The research question in this case study was, “What are the dimensions of the
innovation capability construct for small service businesses?” The innovation capability (IC)
literature describes this construct as the capacity of a firm to develop new products,
processes, and systems (Lawson & Samson, 2001; Prahalad & Hamel, 1990) in order to
compete in dynamic competitive markets. It has been proposed that firms with “good” levels
of IC have a sustained competitive advantage and use it to achieve higher levels of
performance (Sharon A. Alvarez & Barney, 2001). Thus, it is important to understand IC to
assist firms in improving their ability to innovate and hence their abilities to survive, compete
and grow.
The relatively small number of empirical IC studies undertaken to date are primarily
in the manufacturing sector (Guan & Ma, 2003; Yam, Guan, Pun, & Tang, 2004), although
one study has been undertaken across a variety of professional service firms (Bowdle, 2005).
These studies are characterised by researchers developing scales from different theoretical
starting points that result in measures with significant differences in the number and nature of
dimensions and scale items (Balan, 2013; Balan & Lindsay, 2009). For example, one measure
consists of eight dimensions with 101 scale items (Terziovski & Samson, 2007), another
comprises seven dimensions and 70 scale items (Guan & Ma, 2003), a further study includes
two dimensions with 10 scale items (Tuominen & Hyvönen, 2004), and another comprises
one dimension with five scale items (Grawe, Chen, & Daugherty, 2009). The variation in
dimensions and number of scale items used by researchers supports the proposition that
“there is no clear agreement of what the real variables of innovation capability might be”
(Lawson & Samson, 2001, p. 389), and that the nature of innovation capability may depend
on the industry sector (Lawson & Samson, 2001). For these reasons, this research adopted a
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 3
“grounded” approach consistent with the philosophy of pragmatism to investigate the
dimensions of IC for the particular sector being investigated (small service businesses). This
is similar to the approach adopted by Rosas and Camphausen (2007) to develop a scale to
evaluate a particular type of family support program.
The sample and data collection
The hotel sector is a good example of the services sector (Lovelock, 2001; Sundbo,
1997). It is prevalent in communities across most developed countries, has a relatively high
profile, contributes to the economic development of communities, is a significant employer,
is subject to rapid and continuing change, and is highly competitive. This research focused on
independent hotels in Australia that are classified in the industry as “general hotels” or
“pubs”. These constitute the majority of Australian hotels, and are typically independently
owned, or members of small groups, with fewer than 10 full-time employees (ABS, 2006).
Most general hotels are local businesses and draw their clientele from the surrounding
localities. About 50% of these hotels provide accommodation and their facilities are either
ungraded, or range between one and three stars (ABS, 2006). A major feature of general
hotels is that they are small service businesses managed by people with a direct day-to-day
involvement with the business, and who are frequently the owners as well as being the CEOs.
This research excluded four and five-star hotel chains, and two large groups of general hotels,
as innovation decisions in these businesses are largely made in corporate head offices.
A first stage of exploratory qualitative research showed that the “richest” interview
data were obtained from hotels in suburban areas with higher levels of competition and
operated by owner/managers who were known to their industry association to be innovators.
In the subsequent stage of research that is reported in this case, Researcher 1 interviewed six
“innovator” owner/managers as key respondents in each of six Australian capital cities -
Brisbane, Sydney, Melbourne, Hobart, Perth and Darwin - to provide a theoretical sample. A
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 4
relatively unstructured interview questionnaire was used in line with the “grounded”
approach to encourage participants to describe innovation initiatives in their particular
business. The aim was to identify from their narratives as many different factors as possible
that might be relevant in supporting such innovation activities, whether these concerned
innovation in products or services, or in operating systems or methods used in the business.
Interviews were carried out in the premises of each of the participants; the duration of each
was approximately one hour. The researcher took notes, interview recordings were
transcribed by a professional agency, and the researcher checked the transcriptions against
the recordings. Transcriptions were used in each of the two methods of analysis described
below, as shown in Figure 1.
“Grounded” qualitative data (key informant experiences
relating to innovation activities)
Analysis using matrix method (analysis of “grounded” data
using NVivo)
Analysis using network method (analysis of “grounded” data using concept mapping and
UCINET 6.0)
Comparison of “grounded” dimensions of Innovation
Capability
Figure 1: Research overview
Data Analysis Using the Matrix Method with NVivo
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 5
The matrix method of analysing qualitative data (using NVivo) is the one most
familiar to qualitative researchers (Lewis, 2004; MacMillan & Koenig, 2004; Miles &
Huberman, 1994). Researcher 1 analysed the interview data using an inductive method
borrowed from the “grounded” approach (Bringer, Johnston, & Brackenridge, 2006;
Charmaz, 2006; Glaser & Strauss, 1967), and carried this out using NVivo (QSR
International Pty Ltd, 2010).
Drawing from Charmaz (2006), the analysis started with incident to incident coding
(p.53), as the purpose was to discover patterns and contrasts (p.55), as well as fit and
relevance (p.54) of the categories that were identified. In this way, components of innovation
capability were created as categories or NVivo nodes. The analysis started with the Sydney
interviews, as these were judged by Researcher 1 to be the richest, in terms of providing the
greatest number of innovation examples.
During the process of coding data, Researcher 1 named the nodes or categories,
recorded reflections on the incidents and on the meaning of the categories, and their
relationships to other categories in memos, as suggested by Charmaz (2006, p. 72) . This used
the NVivo facility for creating a memo for each individual category, and allowed aspects of
manager perceptions of innovation and innovation capability to emerge in an unprompted
way, thus providing rich results.
Researcher 1 then carried out focused coding (Charmaz 2006, p. 57). This used the
categories identified by coding the Sydney interviews as the basis for coding the other
regions using the same incident to incident coding approach. The purpose was to identify
similarities and differences in those regions, compared with Sydney. In practice, this meant
that further categories (NVivo nodes) were created as needed, whereas some other categories
were not used (to the same extent) in coding interviews for the other locations. In effect, this
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 6
process involved comparing data for the city being analysed, with the Sydney codes to refine
them (Charmaz 2006, p. 60). The analysis was facilitated by using the memos for each
category, and taking into account the observations and reflections that had been recorded
during the activity of coding data into the individual categories. This approach resulted in the
creation of 38 different nodes, together with a further eight nodes derived from participant
statements relating to barriers to innovation, resulting in a total of 46 NVivo nodes or
categories.
During the focused coding process, an exercise was carried out to assess the
feasibility of using two coders to encode this type of data using a grounded approach.
Researcher 1 briefed a third researcher (MacQueen, McLellan, Kay, & Milstein, 1998), and
they independently coded the same two interviews and reviewed and discussed the results. A
further set of six interviews was each coded separately by the two researchers. The average
intercoder reliability, however, was 61%, which fell far below the 70% minimum (Miles &
Huberman, 1994, p. 64), with significant variability in concurrence at the individual category
level. In addition, it was found to be very time-consuming to compare the two outcomes of
coding using NVivo, to try to resolve coding differences. As a result, the attempt to use two
coders was abandoned, and Researcher 1 coded all the interviews.
The 46 categories were then grouped using “axial coding” (Strauss & Corbin, 1998)
to identify themes or dimensions of innovation capability, and this was done in two stages:
Researcher 1 prepared descriptors of each category, and verbatim examples of each
were provided to two other experienced researchers in the field of innovation and
entrepreneurship. These two independent researchers used this information, as well as
comments relating to barriers to innovation, to develop nine separate dimensions that
included the 34 categories.
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 7
Guided by the NVivo coding stripes as described by Bringer et al. (2006, p. 255),
Researcher 1 checked this classification based on a detailed examination of the
extensive content of each category, and adjusted the classification after discussion
with the other researchers.
This constituted an abductive (Reichertz, 2010) or “grounded” approach for
identifying the dimensions and components of innovation capability for this sector ab initio
(Charmaz, 2006; Glaser & Strauss, 1967; Strauss & Corbin, 1998), and it resulted in the
dimensions of innovation capability shown in Table 1.
Table 1: IC Dimensions identified from the matrix analysis
IC Dimensions NVivo Categories/NodesAlliances Alliances with organisations such as external agencies, other hotels
and suppliersCustomer intelligence Customer feedback, customer knowledgeBusiness environment awareness
Awareness of constant change, and awareness of competition, regulations, business trends, market position, technology changes, foresight
Manager characteristics
The manager’s personal knowledge, knowledge about the business, leadership and lifestyle
Experimentation Including pro-activenessHuman resources and human capital
Having good staff, job design, staff incentives and motivation, team culture, team knowledge, formal education, formal skills training, in-house training and organisation structure
Operations Having good operations, management systems and quality controlResource awareness Financial investment and resource managementStrategy and planning Planning, vision, strategic view of the business and portfolio
management
Rigor was achieved in this matrix analysis by ensuring that the data selected was
adequate and appropriate, and that there was a documented audit trail consisting of “raw data,
data reduction and analysis products, data reconstruction and synthesis products, process
notes, materials relating to intentions and dispositions, and instrument development
information”, with the intention that others “can reconstruct the process by which the
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 8
investigators reached their conclusions” (Morse, 1998, p. 77). The audit trail was embedded
in the nodes and memos in the NVivo software (Bringer, Johnston, & Brackenridge, 2004).
Data Analysis Using Concept Mapping
Concept mapping, used generically to describe the representation of ideas in network
form, is also considered a type of integrated mixed method (Greene, Caracelli, & Graham,
1989). The key strength of concept mapping is the integration of qualitative and quantitative
methods, where qualitative information can be represented quantitatively, and quantitative
analysis is enhanced by qualitative judgement (Alvarez & Barney, 2013). Concept mapping
is useful as it helps qualitative researchers capture and discover meaning in social reality
through words and pictures, allowing concepts and ideas to emerge (Rappa, 2001). UCINET
6.0 social network analysis software (Borgatti, Everett, & Freeman, 2002) is used to generate
maps and data displays to represent the relationships between the ideas and illustrate a
conceptual framework, presented as a concept map (Kane & Trochim, 2007). Other software
that performs concept mapping includes Pajek, NodeXL, NetMiner for PCs, as well as Gephi
that is suitable for Apple computers.
The concept mapping method was carried out by Researcher 2, who was not involved
in the original interview process or in the matrix analysis (using NVivo) described above.
Researcher 2 reviewed the 36 transcripts to extract statements provided by participants that
described what was required to be innovative in their businesses, or referred to barriers to
innovation. This process resulted in a total of 377 individual verbatim statements that were
entered into an Excel spreadsheet. After careful review, however, Researcher 2 averaged out
129 duplicate statements (Borgatti et al., 2013, p. 258) to result in item reduction, leaving a
total of 248 statements to be analysed. Researcher 2 then went through a linking process, to
identify which of the 248 statements were most similar to each other. This was done by
identifying keywords from each phrase, and matching these based on the context of the
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 9
statement. Concept mapping requires that this linking process be done as parsimoniously as
possible to ensure the number of linkages of each individual statement is kept to a minimum
(for example, less than five or six linkages) otherwise the map produced by the software will
be too dense, and difficult to interpret. Researcher 2 identified 361 one-way linkages, based
on perceived similarity of statements.
These linkages were entered into an Excel spreadsheet and uploaded to the UCINET
6.0 social network analysis software (Borgatti et al., 2002). Using the NetDraw function, a
three dimensional concept map was created, with each statement represented by a numbered
node on the map. This allows the researcher to return to the original data spreadsheet and
identify exactly where each statement is located on the map, and its relationship to other
statements. Researcher 2 then analysed the resulting concept map using Girvan-Newman sub-
group analysis (Girvan & Newman, 2002). This analysis presumes that within a community
there will be sub-groups, and uses statistics to measure the “between-ness” of the clusters
within the map. This analysis allows the number of clusters to be varied as required by the
researcher, and Researcher 2 evaluated each set of clusters. This sub-group analysis identified
12 clusters; the researcher merged two clusters to reduce these to 11 meaningful clusters.
Fewer clusters did not yield enough variation in the statements, and more clusters did not
provide significant insights. As this method makes it possible for the researcher to track
where each individual piece of data is at any one time within the process, the researcher made
some adjustments to the links where statements did not appear to “fit” within their cluster.
Notes pertaining to the amended links were recorded. Figure 2 shows how the 248
qualitative comments extracted from the 36 in-depth interview data were grouped into 11
clusters or themes to form a concept map.
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 10
Figure 2: IC Dimensions Concept Map
Researcher 2 referred to the original statements grouped together in each cluster to
arrive at names or labels for each theme. The 11 clusters are named and described in Table 2.
Table 2: IC Dimension Concept Map Clusters
Cluster number in Figure 2
Node Shape Cluster Name
1 Circle Multi-skilled and trained staff
2 Hourglass Partnerships and alliances
3 Square Systems and management procedures
4 Down triangle Planning and resources
5 Square Facilities and infrastructure
6 Up triangle Compliance and regulations
7 Circle Customer engagement
8 Down triangle Community engagement
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 11
9 Hourglass Market scanning
10 Circle Objective data and research
11 Diamond (& Cluster
12, Square)
Experimentation
Comparing the Results of the Two Methods of Analysis
At this point, both researchers carefully reviewed the statements in each cluster to
ascertain the appropriateness of the name for each cluster, and compared the findings from
the grounded matrix approach using NVivo, and the concept network approach. The goal was
to identify whether it was possible to align the IC dimensions revealed by each data analysis
method drawn from the same sample of 36 in-depth interview transcripts. As shown in Table
3, as a collaborative exercise the researchers were able to match the nine matrix analysis
dimensions with the 11 concept map dimensions.
Table 3: Comparison of IC dimensions generated from both methods
# Matrix (NVivo) analysis dimensions of IC
Concept mapping dimensions of IC
1 Environmental awareness Objective data and research, Compliance and regulations, Market scanning
2 Alliances Partnerships and alliances3 Customer intelligence Customer engagement, Community
engagement4 Experimentation Experimentation5 Strategy and planning Planning and resources6 Manager attributes Multi-skilled and trained staff7 Human resources and human capital Multi-skilled and trained staff8 Resource awareness Facilities and infrastructure9 Operations Systems and management procedures
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 12
The comparability of the dimensions in Table 3 indicates that the outcomes of the two
separate analyses were quite similar at this broader level, thus indicating that they could be
used interchangeably for the purpose of further research, such as for the development of
items that could be included in a scale for innovation capability.
Discussion
The research question addressed in this case study required an exploratory
“grounded” approach to analyse qualitative in-depth interview data. Although the relevant
literature was reviewed and formed a context for the research, the dimensions of IC in this
research emerged through a comprehensive “grounded” analysis of in-depth interviews with
hotel owner/managers using a matrix method. The same qualitative data set was also used to
generate a concept map using a network method, and similar findings emerged. Importantly,
this exploratory research identified factors not previously identified in IC scales developed
for other industry sectors or for other types of service businesses (Balan, 2013).
With regard to the two methods presented, the analysis of qualitative data can be
described in terms of the stages of coding data into nodes or categories, integrating
categories, developing theoretical insights, and continuing the analysis until saturation is
reached (Shah & Corley, 2006, p. 1828), and these stages provide a framework for comparing
matrix analysis using NVivo, and concept mapping.
Coding data into nodes or categories, and naming categories
Matrix analysis (NVivo) requires the researcher to devise and name nodes, and to
allocate similar data elements to appropriate nodes, and write memos that capture the
researcher’s reflections on the data during this analysis (Bringer et al., 2006; Hutchison,
Johnston, & Breckon, 2010). This is a process that requires clear judgement and consistency,
and some researchers have identified problems in accuracy of data assignment, and incorrect
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 13
labelling of codes (Davis & Meyer, 2009). In a like manner, concept mapping requires the
researcher to identify similarities between data elements in as objective a manner as possible.
The researcher, however, is not required to name categories or nodes during coding, as this
interpretive step is taken only after the concept mapping software has revealed clusters, and
the researcher has examined data elements in those clusters. Similarly, reflection on the data
groupings (categories) takes place at this later stage when the researcher has the benefit of
examining the data arrangements; this replaces the writing of memos.
Integrating categories
Axial coding involves making comparisons at the category and subcategory levels
(Strauss & Corbin, 1998), and is carried out in matrix analysis with the assistance of facilities
such as NVivo coding stripes (Bringer et al., 2006). Concept mapping software generates
cluster maps that directly display the integration of categories for conceptual development
(Figure 2). In particular, the researcher has the ability to select the number of clusters
generated by the software, and examine the data elements contained in clusters of different
sizes to identify possible integrative relationships.
Developing theoretical insights, and providing a visual representation
NVivo provides matrix coding query functions that can be used to explore
relationships between categories at different levels, and relationship nodes can be created to
identify possible relationships between categories. Theoretical development can be further
supported by the modelling tool in NVivo that allows the researcher to manually move nodes
into related clusters and then interrogate nodes to identify the data behind them (Bringer et
al., 2006; Hutchison et al., 2010).
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 14
In comparison, concept mapping generates a visual output drawn directly from all of
the data elements (Figure 2). The software enables two and three-dimensional views; this
provides the qualitative researcher with an efficient, yet very clear and rich, illustration of the
analysed statements, the constructs that emerge from the analysis, and their inter-
relationships (Corley, 2011). This gives the researcher (visual) insights into the structure of
the data and possible theoretical development, and allows the reader to “connect the raw data
with the analyzed data, and the analyzed data with the emergent theorizing” (Bansal &
Corley, 2012, p. 511). For example, in the concept map developed in this research, the cluster
“Location and physical assets” is located next to the cluster “Community engagement and
relevance”. This provides additional insights into the data that can be used for theory
development.
The use of the Girvan-Newman (2002) analysis provides an additional benefit,
making it possible to determine when an optimum number of themes (clusters) has been
reached, by observing if the addition of another cluster adds understanding or insight. In this
case, the researchers considered that 11 clusters were optimal, as when the software was set
to identify 13, 14, and 15 clusters, only one or two individual statements were separated from
the existing clusters each time. In a different qualitative research project, the researchers
determined that only three clusters provided the most useful theoretical insights into the
problem being investigated (Reynolds, Balan, Metcalfe, & Balan-Vnuk, 2014).
Achieving saturation or theoretical density
In the analysis described in this Chapter, data analysis using NVivo was carried out
location by location, and this made it possible to identify when saturation was reached
(Charmaz, 2006, p. 299). The same approach can be applied when analysing data using
concept mapping. In particular, the manual coding process allows additional data to be added
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 15
to an existing data set for cluster analysis. The cluster maps for each increment in data
grouping can be compared, and this will reveal if any new clusters emerge. Saturation is
achieved when the cluster structure does not change, but existing clusters become denser. The
same approach can be used to carry out “coding-on” to develop dense categories and explore
links to other categories (Bringer et al., 2006, p. 255)
Other key differences identified between the two methods included the audit trail,
intercoder reliability, and the time required for analysis.
Audit trail
A central consideration in qualitative analysis is the trustworthiness or conceptual
soundness of the analysis. In particular, trustworthiness has been described as being made up
of the following criteria: credibility, transferability, dependability, and conformability. This is
supported by the provision of a systematic audit trail (Lincoln & Guba, 1985). The need to
have a sound audit trail has been emphasised in the literature (Bowen, 2009; Shah & Corley,
2006). For analyses using NVivo, an audit trail is made up of records of analysis and project
journals including descriptions of analytical procedures (Bringer et al., 2004; Hutchison et al.,
2010). Nonetheless it is time-consuming to read and follow the audit trail for the generation
of nodes and categories using the matrix approach. This makes it challenging for other
researchers to “replicate” a given analysis, or identify possible coding errors (Davis & Meyer,
2009).
In comparison, the concept mapping process provides a clear and straightforward
audit trail, consisting of a limited set of documents: the raw data records, a spreadsheet
displaying individual comments or data elements with similarity codings, and a spreadsheet
with data elements grouped by clusters. This means that one researcher may collect the
statements, make the links between similar statements on a spreadsheet, generate the map,
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 16
and group the data elements in another spreadsheet. Another researcher, by virtue of the
visual output and the spreadsheets, can then examine and question the link between each
statement in the map. The transparency of the method allows other scholars to use the limited
number of documents described above to examine rigorously the steps taken to arrive at a
result, which is a desirable attribute for any research.
An additional benefit of concept mapping is the qualitative evaluation of construct
validity. This is a challenge in qualitative data analysis, as it is not always clear whether
statements placed in the same node or category in fact belong together. In concept mapping,
the items most similar to each other are clustered together, and this allows construct validity
to be more readily assessed, both visually and in reference to the Excel spreadsheet
displaying the grouped data elements. These audit trail documents provide a sound basis for
researcher collaboration.
Intercoder reliability and researcher collaboration
The experiment carried out in this research that used two researchers to independently
analyse the same data using matrix analysis (NVivo), showed that intercoder reliability when
using a grounded research approach was unreliable, falling below the recommended 70%
concurrence (Miles & Huberman, 1994, p. 64). In addition, it was found to be time-
consuming to compare the NVivo coding outcomes for the two researchers. This means that,
in practice, it may not be practical for others to “reconstruct the process by which the
investigators reached their conclusions” (Morse, 1998, p. 77). In the case of concept
mapping, collaboration is facilitated by the existence of the clear audit trail for each step in
the process. The list of statements, ordered by cluster, can be checked against the visual
output and examined carefully to identify any possible anomalies, and these can be quickly
remedied. In this research, Researchers 1 and 2 were able to subsequently review Researcher
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 17
2’s coding and to agree quickly on the optimum coding of statements, using the Excel
spreadsheet records. This demonstrates the value of concept mapping in supporting
researcher collaboration.
Time required for analysis
There was a significant difference in the amount of time required by each method to
analyse the same dataset. The 36 interview transcripts ranged from 7,000 to 22,000 words
each. Implementing the matrix approach using NVivo, many hours were required to go
through each line of each transcript, and identify nodes and categories while becoming
familiar with the data. This was an ongoing process, and Researcher 1 had to constantly
evaluate whether new nodes were required based on new statements, or whether statements
could fit into an existing node or category. When using the concept mapping method,
Researcher 2 identified statements that helped to answer the research question and pasted
these into an Excel spreadsheet. The statement linking process, where each statement was
carefully examined in relation to the remaining statements in the spreadsheet, was time
consuming, although the researcher was not required to identify themes, even though some
could be observed. Categories were identified as clusters in the concept map generated by the
UCINET 6.0 software. Researcher 2 used the map to qualitatively evaluate the homogeneity
of the statements in each cluster. The researchers estimated that the concept mapping process
took considerably less time to execute than the matrix approach using NVivo.
Due to these aspects, it is suggested that concept mapping, a network method, is a
valid and credible approach for analysing this type of qualitative data. It does not depend on
the researcher a priori identifying nodes or categories, it provides a visual output to assist in
theory development, and it provides a clear audit trail, while saving time.
Summary
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 18
The purpose of this research was to make a direct comparison between two different
research methods to determine whether one may have advantages over the other. A
“grounded” investigation of the innovation capability construct was used as a case example.
Data was collected using relatively unstructured in-depth interviews with the owner/managers
of small general hotels in Australia. Interview transcripts were analysed independently; firstly
using traditional matrix analysis and NVivo software, and secondly by implementing concept
mapping using UCINET6.0 software. It was found that both methods produced similar
results. This illustrates the proposition that “there is no single right methodology for
organising and analysing data, but rather a logic in the methods that ties together the research
question, data collection, analysis, and theoretical contribution” (Corley, 2011, p. 236).
Many researchers have found matrix analysis using NVivo (for example) to be a
valuable research tool. This case study suggests that qualitative researchers might consider
adding concept mapping to their repertoire as a valid and credible approach for analysing this
type of qualitative data when using a grounded approach. This exercise found that concept
mapping does not depend on the researcher a priori identifying nodes or categories, so that
data interpretation can be left a later stage of analysis when it is facilitated by a visual output
that helps to identify relationships between categories or nodes. In addition, concept mapping
provides a clear and detailed audit trail that facilitates collaboration and verification.
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 19
Chapter 29: Concept mapping as a methodical and transparent data analysis process
(3) Textbox summarizing the innovation:
Qualitative “Grounded” data requiring analysis
Data analysed using matrix method with NVivo
Data analysed using concept mapping method with UCINET 6.0
Results compared, and concept mapping: • Facilitates incident to incident coding, axial coding and integration of categories•Provides a visual output for developing theoretical insights and identifying saturation•Provides a clear audit trail•Enables researcher collaboration•Saves time
V2-6 6Nov14
Concept mapping as a methodical and transparent data analysis process 20
Table n Use of concept mapping for grounded data analysis in qualitative research
Reference Research Context How Innovation was used
Outcomes/results
(Balan-Vnuk, Dissanyake, & O'Connor, 2014)
Evaluating entrepreneurship policies for the Sri Lankan government
Summarise secondary data to identify themes in publicly available data
Identify eight clusters of development strategies to support entrepreneurship education and new venture formation
(Balan-Vnuk, 2013) Identify business model strategies for non-profit social enterprises
Analyze in-depth interview data to identify business model strategy types
Development of a typology of five business model strategies adopted by non-profit social enterprises in Australia
(Reynolds et al., 2014) (1) Identify the dimensions of innovative business models of general hotels
Analyze qualitative data from the websites of a sample of hotels
Identified three major themes of business model innovation that can be used by hotel managers to improve their business
(Balan & Balan-Vnuk, 2013) Identify the dimensions of student engagement with a particular teaching method
Analyze student engagement with the Team-Based Learning method of teaching, with data obtained from minute paper evaluations
Identified common dimensions of student engagement with the Team-Based Learning method
(Balan, Balan-Vnuk, Lindsay, & Lindsay, 2014)
Identify the dimensions of student learning motivations
Analyze student learning motivations, with data obtained from minute paper evaluations
Identified common dimensions of student learning motivations for six classes