CONCEPT MAPPING AS A RESEARCH TOOL1

Concept Mapping is one of the many research methods or tools that can be used to analyse qualitative data. Others include cluster analysis, ideas networking, multidimensional scaling, argument mapping, Boolean analysis, Leximancer, network analysis, content analysis, and nVivo.

Methods like content analysis, cluster analysis, and multidimensional scaling offer an objective epistemology with software mechanically comparing text strings and producing simple graphics, counts or tables as output. These methods can be used to analyse data consisting of either long strings of qualitative data (such as transcriptions of in-depth interviews), or data consisting of short comments. The following flow chart (from Mike Metcalfe), illustrates the rationale for selecting Concept Mapping (or Idea Networking) as a method of analysis:

A CASE EXAMPLE: INVESTIGATING STUDENT ENGAGEMENT

As in any research situation, the research method and tool are selected that will best address the research question. In this illustration of the use of Concept Mapping, the research idea was to find out in what ways undergraduate students are engaged by Team-Based Learning as a particular teaching method. The researcher wished to implement a grounded approach to explore the themes or concepts relating to student engagement in this context.

The specific research question was: "what are the dimensions of student engagement in Team-Based Learning in an undergraduate entrepreneurship course?"

(This example is drawn from a refereed conference presentation:Balan, P & Balan-Vnuk, E 2012, 'Student engagement with Team-Based Learning in undergraduate entrepreneurship courses: a pilot study', Education Research Group Adelaide, 19-21 September 2012, Adelaide, Australia.)

1. WHY USE CONCEPT MAPPING?

Concept mapping can be described as an inductive means of clustering similar activity statements using the clustering function in network diagramming tools. This method is used for a range of reasons:

  • it is appropriate for the research question; "what are the underlying dimensions of engagement?"
  • it is straightforward to implement with the type of data that is collected (short statements)
  • note that Concept Mapping can be used effectively as a tool to analyse lengthy in-depth interviews
  • it is a relatively quick method to implement in this situation where data is collected and processed, and results reported to the classto give students feedback
  • it can produce meaningful/useful results from large or small data sets (from 30 qualitative statements through to 120 or more)
  • it produces a graphical output that identifies clusters of concepts or themes and provides output that is relatively easy to explain
  • clusters can be explored at different levels of detail
  • the graphical output shows the relationships between clusters and this helps to identify relationships between the underlying concepts or themes
  • the UCINET6 softwareis inexpensive (U$60), and there is good explanatory information on the provider's website:
  • importantly:
  • the process of implementing concept mapping as described provides an audit trail that records each step of analysis. This allows analysis to be assessed and critiqued, and also allows collaborative analysis
  • although the maps produced in this process depend on the keywords that are identified and linked, another researcher should arrive at the same results using the same keywords (thus allowing replication)

2: HOW WAS CONCEPT MAPPING IMPLEMENTED?

STEP 1: Data was collected

This example usedundergraduate students taking an entrepreneurship course. At an appropriate stage in course delivery, students were given a blank sheet of paper, and were told to sit apart from others. They were asked to write down one or two ways in which they found a particular teaching approach (Team-Based Learning)to be engaging. The sheets of paper with these unprompted student comments were collected in a manner to ensure that participation was both voluntary and anonymous. These comments (examples shown below) provided the raw data for this analysis.

STEP 2: Data collation and coding

In this example, the data was entered verbatim into an Excel worksheet.

Datawas coded (manually) as follows:

  • the researcher started with the first statement, and identified the keywords in that statement
  • the researcher then identified the numbers of the following statements, that were close to or identical in meaning, using the keywords as a guide.The statement number was recorded as shown in the following figure (eg items 13, 15, 18 in the figure below were considered to be similar to item 1)
  • it is important at this stage not to try to interpret the statements (interpretation comes later). In other words, each statement was taken at face value so that this was a fairly mechanical process. It is considered desirable to have six or fewer statements coded to the statement being examined.This condition can be hard to achieve with a large dataset.
  • after the first statement was coded, the researcher moved to the second statement, and repeated the process working down the list. In other words, each statement was coded against the statements below it on the list. The further down the list, the less the possible number of statements that could be coded.

The following figure shows an example of the coding for this dataset.

STEP 3:Coding data was entered into UCINET6

Coding datawas entered into an Excel-like matrix in the UCINET6 software. This was done by entering a "1" into the cell that was the intersection between each of the comments as coded, and as shown in the figure.

STEP 4: Concept maps were produced

UCINET6 performs what is in effect a factor analysis of the data. Clusters are formed because this software has a clustering algorithm called “spring embeddedness” that makes the nodes repel each other and encourages clusters to form. The output is a set of clusters of data items, and each cluster can be described as a “pragmatic concept”.

The nodes in these maps are the statements included in the analysis. The number for each node on the map is the number of the statement in the dataset. The lines are the links between the statements, and have equal length as these are actually three-dimensional maps.

Thisprocess produced maps with a number of clusters determined by the researcher.

For example, three clusters could be identified, but the researcher selected what appeared to be an "optimal" number of clusters; e.g. when additional data groups did not seem to add to the overall "picture" ("saturation" appeared to have been reached).

In this case, 10 clusters appeared to be optimal, as shown below

STEP 5: Themes were identified

The statements that made up the raw data (in the Excel spreadsheet) were then sorted into the groups or “pragmatic concepts” shown in the cluster maps. Visual inspection of the resulting collections of statements allowed the researcher to:

  • check the consistency with which similar statements were grouped into the same cluster (this revealed mis-coded data that was corrected).
  • reflect on isolates; it was possible to recode these into one of the clusters/themes.
  • the cluster maps needed to be redrawn after any changes in coding.
  • the aim was to arrive at groups of items that were as homogenous as possible. This was done in discussion with a collaborator

The researcher gave a cluster of statements a meta name, or a collective name that is called a concept (because it is a conceptual idea) that was drawn from the elements in each cluster (as shown below). This was also done in discussion with a collaborator.

STEP 6: Discussion of results

These maps were the start of analysis and interpretation, as the results were related to the theory in this field.

3: CONCLUSION

This example shows that concept mapping is a useful method to:

  • analyse a dataset of qualitative statements to extract themes
  • analyse data at different levels of detail. This is done by implementing different levels of clustering that allow the researcher to examine sub-clusters and thus gain insights into the structure of the data
  • analyse data in a manner that can lead to theory building. This can be done by exploring the relationships between clusters to gain insights into the relationships between concepts or themes.

ACKNOWLEDGEMENT: Mike Metcalfe (School of Management, UniSA) provided comments and suggestions on an early version of this document.

Peter Balan, MGN School, UniSA, , 24 October 2012