Researchers analyzing qualitative data, such as interviews or written statements, often apply codes to these texts as a way to label and organize concepts and themes found in the data. These codes become the basis of analysis and the way in which researchers both understand and explain what is occurring in the data. Determining how important a particular code is compared to the others and how the codes relate to one another is typically limited to tallying the number of times a code occurs (frequency counts) or instances when codes overlap with one another (code co-occurrence). Presenting and communicating findings from the research is therefore also limited to reporting these same measures, along with providing excerpts from the data that correspond to the codes of interest. Visual representations are rare due to the verbal nature of the data. In this article, we create networks using the chronological location of the codes as they were applied to the text, resulting in a visualization that illustrates the interrelations of the codes in the data. By applying methods from network analysis, additional measures reflecting the relative importance of the codes to one another can be extracted from the networks and illustrated visually in the network graphs. Using this method of analysis can help researchers better understand the relationship of all codes applied to a data set, and offers an additional way to communicate one’s analyses and findings in the form of a network visualization.
This article came out of my postdoc at the Saron Lab at UC Davis where I worked with Clifford Saron and Jennifer Pokorny. While there I was part of the Shamatha Project; a large-scale examination of the longitudinal effects of an intensive 3-month meditation training retreat on aspects such as attention, cognition, emotion regulation and health. The qualitative data I was working with came from semi-structured interviews were conducted with the participants (N=33) at three time points in the retreat (it was, to put it mildly, fascinating!).
I find this method really helpful for thinking and for showing others what I think about my coding of an qualitative dataset (interviews, long-form text, etc.). I’m planning to write a walkthrough blog post about how to use this method to help with both determining the relative importance of codes in your code dataset, and for creating a visual depiction of your coding choices. Stay tuned!