Visualizing Relationships in Networks
In this article you learn about the various applications of ATLAS.ti networks and two FAQs are answered: Can I create case-specific networks, and can networks help me to organize codes in a hierarchical manner?
Written by: Dr. Susanne Friese
Situating Networks in the Research Process
The ATLAS.ti network function is a tool that allows you to explore your data visually. When you begin your analysis journey, you explore your data by noticing interesting things. After a while, you are able to label what you notice, and you gain a better understanding of the data during the process of coding (see chapter 5 in the book: Qualitative Data Analysis with ATLAS.ti). At first, this is a descriptive understanding. With a prolonged stay and further exploration, you are able to describe the various aspects of your data and their specifications in the form of a well-developed code system. This enables you to dig a bit deeper and to ask more specific questions utilizing several different tools provided by the ATLAS.ti workbench.
As you go through your research questions step by step in the process of further analysis, writing memos is a must. By writing, you put into words the results you see in the form of tables, numbers and data, and add your thoughts, ideas and interpretations. It is through this process that analysis ‘happens’. As Freeman (2017: 4) puts it: it is important to understand that ‘writing is inseparable from analysis’. It is also during this process that you realize how the different aspects of your data relate to each other. The visualization of these relationships in the form of ATLAS.ti networks is a next logical step (see chapter 7 in the book: Qualitative Data Analysis with ATLAS.ti). It will advance your analysis on another level.
Graphical illustrations enable a different kind of exploration. Images activate different parts of the brain from words and lead to different processing modes (e.g., Khateb et al., 2002). The ATLAS.ti networks support creativity and help in the detailing of an idea or by developing a line of reasoning. They improve metacognition by encouraging a different way of thinking. At the receiving end, they help create a common understanding and help communicate complex ideas and arguments (Freeman, 2017; Novak and Cañas, 2006; Novak and Gowin, 2002). In the following you find some ideas regarding how to use networks in the research process.
Messy Networks to Develop a Storyline
In developing a sample project demonstrating a grounded theory analysis based on the Strauss/Corbin approach (Friese 2016, Friese 2019b), I came to the point where I had to choose a core category. The study was about war experiences of veterans. I decided that my analysis would be based on the concept of ‘coming home’ after the war. At the time, I already had some ideas on how the ‘coming home’ code could be linked to other concepts and categories.
So, I created a network and pulled in my ‘Coming Home’ code. Based on a number of questions I asked myself, I added more codes to the network. What factors make a home-comer feel that he or she has arrived home? What hinders a successful return? Which strategies work against this? Are there differences between those who fought at the front and those who served behind the lines, like the paramedics? Which coping strategies were used during and after the war? What was the original attitude towards the war? Has this attitude changed? In other words, I worked backward in time from the present to the past to develop my storyline.
Figure 1 shows the network I created. It looks a bit messy. But the aim was not to create a network that can be used in a presentation or report. It was just for me and helped me think about possible connections by placing the nodes in the network, creating links and naming relations. Based on this large network, I focused on partial aspects of the story and created several smaller (less chaotic) networks.
At the same time, I wrote down my thoughts, ideas and interpretations in research-question memos. As mentioned above, writing memos is also an integral part of this analytical phase. Creating networks helps you see things in the data, but you must write them down, otherwise the ideas that come to your mind will be fleeting and forgotten tomorrow. And as you write, the connections become clearer to you, and gradually the individual parts can be linked and result in a coherent story.
Using Networks to Discuss Findings with your Adviser or Colleagues
You may send your adviser or colleagues an excerpt from your analysis chapter before scheduling a meeting, so they can get an idea about your work. Prepare one or two networks that visualize the main arguments that you want to discuss and bring a printout or your laptop to the next meeting. Instead of going through pages of text, you can explain your ideas while looking at and discussing the network(s). If there are questions related to the underlying data, you can open the network(s) in ATLAS.ti on your laptop and access the data from within the network.
Talking about the findings based on the printed-out network might, however, already be enough. Working with a different medium, paper in this case, can be a nice change, especially when it comes to visualizing ideas. You can extend your ideas by scribbling notes on the printout or by drawing new ‘networks’ on paper. After the meeting, you can transfer your ideas, notes and paper-based networks to ATLAS.ti. As the data in ATLAS.ti are only a few mouse clicks away, you can verify whether the ideas are valid and still hold when checking them against the data. If so, you can refine your current networks and the analysis you wrote in your research-question memos.
Using Networks to Present Findings
Figure 2 shows the results of a study comparing media reports from 2008 and 2009 on the financial crisis and illustrates how you can make your data come alive in presentations. This requires you to run ATLAS.ti in the background or, alternatively, to use ATLAS.ti to present your findings. Note that the study was developed as a sample study based on a small data set. Therefore, the results are fictitious.
The network below shows factors that have been mentioned by various sources (personal experiences, statistical figures, news agencies and political opinion) as immediate, long-term and individual consequences. Also shown is an activated audio quotation that can be played. Text quotations can be viewed as a preview or in the context of the data. Video quotations can also be played.
For smaller projects, like a master’s thesis, it might be possible to integrate all findings into just one network, but sometimes several are needed.
Using Networks in Presentations and Publications
The following two networks are included in genuine publications and represent central findings. Figure 3 shows one result of my dissertation research illustrating the phases of an addictive buying experience. The original network published in my dissertation (Friese, 2000) looked a bit different as several options were not yet available in version 4 of ATLAS.ti.
For the next example I am indebted to Eddie Hartmann for allowing me to use his data material (Hartmann, 2011). He conducted 20 interviews and developed a case structure for each person based on four main criteria: negation, affirmation, rejective negation and positive substitution. Figure 4 shows two of the cases. One can see overall which criteria applied, and within each criteria, which subcategories were relevant. ‘Affirmation’, for instance, did not apply to case 3 at all, but was strongly present in case 17.
In the next section I discuss two FAQs:
- Can I create a network for just one case / one interviewee / a subset of my data?
- Can I use the networks to organize my codes in a hierarchical manner?
Case-based networks: Can I Create a Network for Each Interviewee?
When you link two codes to each other in one network, the link is not just a private link for this one view. It is used for these two codes in the entire project. Thus, if you pull the two codes into another network, the link between these codes immediately becomes visible. So what can you do if you want to link Code A with Code B using the relation ‘is a consequence of’ in a network for Paul, but need to use the label ‘is a prerequisite of’ in the network for Mary?
The solution is to add more information to the network. If A is a consequence of B for Paul, there is probably a reason for this that you know from reading the data. If A is a prerequisite for B in the case of Mary, there are likely to be circumstances that explain this relation, too. This information is what you need to add to the network. Thus, it becomes even more meaningful. If the reason that applies to Paul and the circumstances relevant to Mary are not yet covered by existing codes, you need to create them. These codes, then, will not contain data; you use them as modifiers to show the kinds of relations that exist in your data. The term for such codes in ATLAS.ti is abstract codes. The frequency is zero and the density is at least one or higher.
Creating abstract codes occurs quite regularly when creating networks. You often need to add codes at this stage to let the network visualize the story you want to tell. Codes have the advantage over memos that you can link them via first-class (i.e., named) relations.
The network in Figure 5 contains one abstract code: ‘comments on blogs’. It stands for a document group. By creating a code (node) instead, representing all blog comments in documents 3 and 5, codes can now be linked using named relations. This would have otherwise not been possible, as connections between documents and codes are second-class links and therefore cannot be named.
Can I Use Networks to Create a Code Hierarchy?
I observe from time to time that researchers use the network function to sort codes into higher and lower order codes. The danger of this is that the code system remains messy and the different levels of analysis get mixed up. If you use the Code Manager to represent the various levels of your codes instead, the network function is still at your disposal for higher-order conceptual-level analysis.
Just as a reminder, the best way to sort and organize codes in ATLAS.ti is by using prefixes as shown in Figure 6 below. You can read more about it in the following article: How to make the best of codes in ATLAS.ti
According to Guest et al. (2012: 72), ‘Hierarchical relationships are probably the most common but are not necessarily the best organizational choice for all the elements in a codebook.’ They only work well if codes can be clearly distinguished in several levels of detail (see also Bazeley, 2007; Richards, 2009; Richards and Richards, 1995). With a few exceptions, most qualitative data analysis codes are best summarized in a flat structure as items that are similar or belong together. For this you can use prefixes in ATLAS.ti as shown above.
If you were to use the network function to link higher- and lower-order codes in a hierarchical manner, you could view a hierarchical structure in the Code Forest (Figure 7). This however only works if you use transitive relationships for linking the codes. If symmetric (undirected) relations are used, codes point reciprocally to each other.
In the Windows version, you find the code forest option when you click on the Navigator button in the Home ribbon. In the Mac version, select Codes > Show Code Forest from the main menu.
What you represent with such a structure is a first phase descriptive analysis. Also note that each code appears on the first level once. So the code forest is not a strict hierarchical display.
The conceptual analysis of the second phase is about linking data across categories (see Figure 8). The aim of using networks for conceptual-level analysis is to illustrate patterns in your data, answers to research questions, models or theories rather than just displaying the categories that describe your data. Such networks are likely to contain subcategories of a variety of categories but only those that are relevant for the context represented by the network.
For a cleaner methodological approach, the recommendation is that you define and visualize higher- and lower-order codes in the Code Manager as explained using capital and lower-case letters and prefixes, and then reserve the network function for advanced conceptual-level work, when you begin to find relations between codes and across categories.
Bazeley, Pat (2007). Qualitative Data Analysis with NVivo. London: Sage Publications.
Freeman, Melissa (2017). Modes of Thinking for Qualitative Data Analysis. NY: Routledge.
Friese, Susanne (2019a). Qualitative Data Analysis with ATLAS.ti (3rd ed.). London: Sage Publications.
Friese, Susanne (2019b). Grounded Theory Analysis and CAQDAS: A happy pairing or remodeling GT to QDA? In: Antony Bryant and Kathy Charmaz (eds.). The SAGE Handbook of Current Developments in Grounded Theory. Chapter 14. London: Sage Publications.
Friese, S. (2016). CAQDAS and Grounded Theory Analysis. MMG Working Paper 16-07
October 2016. (MMG Working Papers Print). https://www.mmg.mpg.de/61773/WP_16-07_Friese-Theory-Analysis.pdf
Friese, Susanne (2000). Self-concept and Identity in a Consumer Society: Aspects of Symbolic Product Meaning. Marburg: Tectum.
Guest, Greg, Kathleen M. MacQueen, and Emily E. Namey (2012). Applied Thematic Analysis. Los Angeles: Sage Publications.
Hartmann, Eddie (2011). Strategien des Gegenhandelns: Zur Soziodynamik symbolischer Kämpfe um Zugehörigkeit. Konstanz: UVK.
Khateb, A., Pegna, A.J., Michel, C.M., Landis, T. and Annoni, J.M. (2002). Dynamics of brain activation during an explicit word and image recognition task: an electrophysiological study – PubMed (nih.gov), Brain Topography, Spring 14(3): 197–213.
Novak, Josef D. and Cañas, Alberto J. (2006). The theory underlying concept maps and how to construct and use them. Institute for Human and Machine Cognition. http://en.wikipedia.org/wiki/Concept_map (accessed 11 December 2020).
Novak, Josef D. and Gowin, D. Bob (2002). Learning How to Learn. New York: Cambridge University Press. (First published in 1984.)
Richards, Lyn (2014). Handling qualitative data: a practical guide (3rd ed.) London: Sage Publication.
Richards, Tom and Richards, Lyn (1995). Using hierarchical categories in qualitative data analysis, in U. Kelle (ed.) Computer-Aided Qualitative Data Analysis. London: Sage Publications, p. 80-95.
Friese. S (2020).Visualizing Relationships in Networks. Retrieved from https://atlasti.com/2020/12/11/visualizingrelationshipswithnetworks
About the author
Dr. Susanne Friese
Dr. Susanne Friese started working with computer software for qualitative data analysis in 1992. Her initial contact with CAQDAS tools was from 1992 to 1994, as she was employed at Qualis Research in the USA. In following years, she worked with the CAQDAS Project in England (1994 – 1996), where she taught classes on The Ethnograph and Nud*ist (today NVivo). Two additional software programs, MAXQDA and ATLAS.ti, followed shortly. Susanne has accompanied numerous projects around the world in a consulting capacity, authored didactic materials and is the author to the ATLAS.ti User’s Manual, sample projects and other documentations. The third edition of her book “Qualitative Data Analysis with ATLAS.ti” was published in 2019 with SAGE publications. Susanne’s academic home is the Max Planck Institute for the Study of Religious and Ethnic Diversity in Göttingen (Germany), where she pursues her methodological interest.