Integrating Social Network and Reflexive Thematic Analysis in Educational Research

Abstract

Social network analysis is widely used to explore the relationships between entities in education to understand how these relationships shape interaction and development. However, established analytical methods stress the processing of quantitative data and provide inadequate guidance regarding the analysis of qualitative data collected through interviews. This article illustrates the integration of quantitative and qualitative analyses by (1) employing three analytical methods to analyse different types of data about an egocentric network obtained from an interview, (2) developing the interpretations of one type of data to influence the analysis of another type of data, and (3) using the results of the subsequent analysis to explain and expand the results obtained in the prior analytical stage. This article further discusses how these strategies can help researchers demonstrate reflexivity and improve the transparency of analysis. It also provides a means to explore and describe the contexts in which learners and teachers are situated when applying social network analysis in educational research.

Plain Language Summary

This article explores how three different methods of analysis can be used together to study social networks. These methods are quantitative social network analysis, qualitative structural analysis, and reflexive thematic analysis. The study uses different types of data about a research participant's social network and applies each method to one type of data. The findings from one type of data are used to inform the analysis of another type, and the results from each stage of analysis are used to explain and expand on the previous findings. This approach helps researchers to be more reflective and transparent in their work, improving the quality of their analysis. It also provides a way for qualitative researchers to better understand and describe the contexts in which their research participants live.

Introduction

Social network analysis constructs individuals as connected and mutually influenced actors in the social world, and it focuses on analysing the network formed by these social actors (e.g., Blaylock et al., 2018; Crossley et al., 2015; Witt et al., 2024). The network analysed may encompass multiple social actors in a bounded system (e.g., a large network composed of mutually acquainted students) or may be a personal ego-centric network (e.g., a network formed by the friends of a teacher; Crossley, 2016; Edwards, 2010). It has been used to approach education-related topics such as the language development of international students (Gautier & Chevrot, 2015), students’ friendship (Blaylock et al., 2018), classroom interactions (Bokhove, 2018; Witt et al., 2024), and teachers’ professional development (Baker-Doyle, 2015; Penuel et al., 2009). This approach also resonates with the general interest in interactions and relationships in qualitative research (e.g., Tracy, 2020). For instance, qualitative researchers may conduct interviews assisted by network graphs such as concentric circles to collect personal egocentric network (ego-net) data about an individual (the ego) and social actors connecting to the ego (e.g., Altissimo, 2016). This method has also helped educational researchers to understand how interactions with social actors can be leveraged to address challenges regarding learning and teaching from a relational perspective (e.g., W. Yang et al., 2021).

Although concentric circles are a cost-efficient tool to elicit ego-net data in qualitative interviews (Van Waes & Van Den Bossche, 2020), limited examples can be found regarding how to unpack the information embedded in ego-nets and to expand the analysis beyond network features (e.g., Altissimo, 2016; Herz et al., 2015). Notably, such network graphs originate from social network analysis, which is an inherently mixed method that collects and analyses numerical, graphical, and textual data (Bazeley, 2012; Yousefi Nooraie et al., 2020). Hence, we aim to exemplify an integrative analysis procedure performed on numerical data embodied in network graphs, graphical data displayed as graphs, and textual data generated in interviews. This procedure consists of three connected analytical methods: well-established quantitative indicators, qualitative structural analysis, and reflexive thematic analysis. The results of two analytical methods associated with social network analysis are developed into inputs for a single process of reflexive thematic analysis, and thus woven together in the analysis outcomes. This article demonstrates the possibility of enhancing the interdependence of methods in data analysis (Bazeley, 2024). It also delineates a way for educational researchers to shift their attention between network features and meanings (Altissimo, 2016), thus facilitating the exploration of interactions within ego-nets and their influences on individuals’ experiences.

From Quantitative to Qualitative Social Network Analysis

Quantitative indicators can serve as “mathematical descriptions and analyses of interactions, relations, and network structures” (Hollstein, 2014, p. 10) to succinctly describe the features of ego-nets, although these indicators are usually used to analyse survey data collected within a bounded system (Crossley et al., 2015; Edwards, 2010). For example, researchers can understand the size of an ego-net by summing the total number of social actors, evaluate the mean closeness of relationships between a focal participant and other social actors, calculate the density of a network by computing the ratio of existing relationships to total possible relationships, and examine the diversity in specific attributes of relationships and social actors in a network (for more details about analysing ego-nets, see Crossley et al., 2015).

Qualitative researchers may transform graphical data into numbers to allow ego-nets to be described with quantitative indicators. A common approach is to use network graphs (e.g., a set of concentric circles with the participant at the centre) as visual aids to assist interviews (Crossley et al., 2015). By inviting participants to place the social actors they identify in concentric circles, researchers can collect graphical data (the participant-generated network graph) and textual data (interview transcript) concurrently (Figure 1). Researchers may extract the numerical data embedded in graphical data and transform it into statistical descriptions of ego-nets. For example, Van Waes and Van Den Bossche (2020) note that researchers can quantify the distance between each layer in a set of concentric circles to capture the degree of closeness between the identified social actors and the participant.

Figure 1.

Ego-net generated by a research participant.

In contrast to the detailed guidance available on quantitative analysis of network data, researchers face a shortage of step-by-step instructions on qualitative analysis of network data generated concurrently with interviews. Although Hollstein (2024) mentions that many researchers employ grounded theory when analysing networks qualitatively, published studies usually do not share enough details about the analytical procedures used to enable others to replicate those procedures, regardless of which method is applied (e.g., Dobbie et al., 2018; Ryan, 2021; Sommer & Gamper, 2018). Researchers may also find it difficult to apply specific strategies to the graphical data collected during qualitative interviews. For example, participants of qualitative interviews may be mutually unacquainted, so it is unlikely for them to nominate other participants when generating social graphs. The consequence is that researchers are unable to create and analyse a whole network by collating the ego-net of each participant, which makes some documented analytical strategies (e.g., Mamas, 2018) inapplicable. Therefore, despite the applicability of some quantitative indicators to analysing the general features of an ego-net (see Crossley et al., 2015), researchers may need to find a complementary way to examine the specific features of participant-generated ego-net graphs.

Qualitative Structural Analysis as a Complementary Method

Qualitative structural analysis, proposed by Herz et al. (2015), outlines steps qualitative researchers might take to examine network graphs in detail, which was previously achieved through quantitative social network analysis (Froehlich, 2020). This method also allows qualitative researchers to provide explanations for quantitative indicators that describe the features of an entire ego-net, as shown in Table 1.

Table 1.

Complementing Quantitative Social Network Analysis with Qualitative Structural Analysis.

Elements	Quantitative indicators of social network features	Foci of qualitative structural analysis
Structure	Size: the total number of social actors
	Density: the ratio between existing relationships and total possible relationships	Clustering: interconnected social actors form clusters that increase density in an ego-net
		Structural holes: the vacuums caused by absent relationships, which makes clusters observable and may reduce the density of a network
Actors and Relationships	Tie central tendency: the mean closeness of relationships between the nominated social actors and a focal participant	Tie strength: the closeness of relationships between each social actor and the participant
	Tie dispersion: an indicator describing the diverse role relationships in an ego-net	Attributes of actors: the roles of social actors in the participant’s ego-net
		Multiplexity of role relationships: one social actor plays multiple roles

Note. Definitions of concepts were adapted from Crossley et al.’s (2015) book and Herz et al.’s (2015) article for illustration. These narrow and simplified definitions will be used in the example presented later in this article.

At the core of this analysis method are the three sets of questions used to examine the structure of an ego-net, the relationships within it, and individual actors’ attributes during the analysis of network graphs and textual data obtained from interviews (Herz et al., 2015; Herz & Altissimo, 2021). By examining these three sets of phenomena, researchers can observe the entire ego-net graph and then examine the structure of the network it represents, such as how social actors form clusters, and the vacuum caused by absent relationships between social actors. Researchers can also examine the social actors who bridge the gaps between clusters and employ these descriptive analyses to explain the quantitative indicator of network density. Qualitative structural analysis allows researchers to examine the closeness of the relationship between a social actor and a participant (i.e., tie strength). Thereby, researchers can shift their attention from the analysis of tie central tendency – a quantitative indicator of the mean tie strength – to the exploration of specific relationships that demonstrate certain attributes. Similarly, researchers can focus on the attributes of social actors (e.g., the roles played by each social actor) and examine the multiple roles played by the same social actor, which enriches the interpretation of tie dispersion – an indicator of the diversity in social actors’ role relationships. Since “meanings and narrative interpretations are closely knitted within the numerical presentation of quantitative results” (Yousefi Nooraie et al., 2020, p. 111), qualitative structural analysis can add to the narrative interpretations of quantitative indicators.

Although qualitative structural analysis may be employed to complement quantitative social network analysis, explanations of how to perform this method are scarce. Froehlich (2020) states that researchers may integrate qualitative structural analysis with quantitative social network analysis by collecting broader quantitative data on a whole network before collecting qualitative data on ego-nets. However, such integration is impractical when researchers are unable to construct a whole network based on the ego-nets of participants who do not know each other. Furthermore, Herz et al. (2015) have highlighted a difficulty in employing qualitative structural analysis in studies indirectly exploring networks, implying a need for researchers to employ other qualitative analytical methods to delve deep into the textual data and understand nuances indirectly related to network features.

Potentials of Integrating Earlier Analyses into Reflexive Thematic Analysis

Through reflexive thematic analysis, it is possible to address problems that qualitative structural analysis cannot solve. Since reflexive thematic analysis has been developed to facilitate the interpretation of qualitative data “centered on exploring participants’ experiences and sense-making” (Braun & Clarke, 2022, p. 10), it helps researchers further their investigation of participants’ sense-making of interactions within a social context, rather than limiting the analysis to network features and participants’ experiences about the entities in ego-nets. Reflexive thematic analysis can also be a pragmatic choice for researchers when analysing textual data. With only six steps, it embodies the essence of abstracting qualitative data into smaller codes and constructing themes that capture broader patterns (for details, see Braun & Clarke, 2021; Terry et al., 2017). Researchers start by familiarising themselves with the data (Step 1) and then develop codes (Step 2) to construct candidate themes (Step 3); they engage in an iterative process of revising and refining codes and themes throughout Step 4 and 5, completing the analytical process by naming the themes (Step 5) and finalising the reports (Step 6).

Braun and Clarke (2021, 2022) argue that reflexivity and interpretive depth should be the aims of good reflexive thematic analysis, and that the quality of reflexive thematic analysis depends more on researchers’ engagement with the data than on criteria that incorporate quantitative perspectives, such as agreement between multiple coders. In other words, good reflexive thematic analysis emphasises not only the outcomes of the analysis, but also the analytical process itself. Specifically, being reflexive requires researchers to consider their own assumptions, experiences, and methodology, and integrate “[e]xisting theories, concepts, and knowledge” (Braun & Clarke, 2022, p. 10) as resources during the analytical process. Furthermore, researchers should demonstrate reflexivity by being more transparent about how they make decisions and how those decisions shape the analysis (Braun & Clarke, 2019).

Reflexive thematic analysis has certain methodological features that facilitate its integration with other analytical methods. First, it recognises that researchers’ assumptions and theoretical understanding shape their research (Braun & Clarke, 2022), which implies that researchers should acknowledge the influence of prior results such as quantitative indicators on their subsequent analyses. Second, reflexive thematic analysis allows researchers to construct knowledge both deductively from existing theories and inductively from textual data (Braun & Clarke, 2024; Fereday & Muir-Cochrane, 2006). This approach suggests that researchers may develop their analysis of numerical and graphical data into an input for the further analysis of textual data. Third, using results obtained from the earlier analytical stage as input may help researchers generate analytical ideas at the initial stage of reflexive thematic analysis. Discussing the genesis of these analytical ideas may increase the transparency of the analytical process (Tuval-Mashiach, 2017), which is a key criterion of rigour in qualitative research (Nowell et al., 2017; Tracy, 2020). Considering these potential benefits, the following example will demonstrate how the analyses and interpretations of ego-net features can be integrated into the analytical process of reflexive thematic analysis.

Integration of Analytical Methods: A Worked Example

The data in this article was extracted from a multiple case study that the first author conducted for her PhD thesis (Figure 2). Although the first author recruited 35 participants (cases) through snowball sampling, this article emphasises the presentation of integration in the analytical process with details from one participant to show that these strategies can still apply to single case studies. The data collection process was approved by the Human Research Ethics Advisory Panel at the authors’ affiliated institution. Written informed consent was obtained from participants prior to the commencement of the study. To minimise risk and protect confidentiality, participants were encouraged to use pseudonyms for themselves and other social actors and were informed of their right to decline to answer any questions and to withdraw from the study at any time. The potential discomfort associated with participation was minimal and primarily related to self-reflection on participants’ careers and social interactions. These risks were considered to be outweighed by the anticipated benefits of the study, including its contribution to supporting teacher education and advancing educational research. Participants also benefitted from opportunities for structured self-reflection on their social networks and future career development.

Figure 2.

Snapshot of the project.

The data collected from Janine, a female teacher working in an international school in a metropolis in Eastern China, was chosen as an illustrative example of the analysis. Janine is the pseudonym that we use to identify the participant in the study (Wang et al., 2026). A constructivist paradigm enabled the first author to investigate multiple realities as constructions of personal experiences and interpretations. The project aimed to understand how resources were exchanged within the ego-nets of returnees with overseas degrees in education (“returnees”) and the influences of such exchanges on returnees’ careers.

Data Collection

Lasting approximately 110 min, the interview with Janine was conducted on an online meeting platform and produced the network graph shown earlier in Figure 1 and approximately 28,000 Chinese characters of textual data. The first author elicited information about the closeness of relationships because the Chinese categorisation of interpersonal relationships emphasises the concept of closeness (Ye, 2004). Janine was encouraged to use colours to differentiate actors belonging to different life domains (e.g., workplace, family, and former schools). The interview questions elicited information about the actors Janine nominated, narratives of Janine’s interactions (e.g., resources exchanges) and relationships with them, and how they had influenced Janine’s career. The first author also prepared prompts to explore network features, including questions about whether the nominated actors knew each other.

Janine generated an ego-net on a set of seven concentric circles. Although Ryan et al. (2014) argue that the seven concentric circles used in their study might have given participants too many choices, we chose a seven-layer design because the careful gradation of relationships this design allows is supported by existing Chinese studies. For example, relationships can be classified, according to their closeness, into kinship (family members), pseudo-kinship (best friends), and friends and acquaintances (Bian, 2018). Chinese villagers are known to classify friends into two levels and acquaintances into three levels, including close acquaintances, acquaintances, and “nodding acquaintances” (Y. Yang, 2009, p. 57). Based on existing literature, the concentric circles in this study categorised friends into best friends, good friends, and normal friends, while acquaintances were classified into close acquaintances, acquaintances, and strangers (e.g., nodding acquaintances). Participants could classify social actors into any of these layers based on their understanding of the closeness of their relationships, placing the closest in the innermost circle and the least close in the outmost circle.

Purposes of Using Three Analytical Methods

Using network graphs during interviews enabled the co-construction of participants’ ego-nets by participants and interviewers (Ryan & D’Angelo, 2018). The numerical data embodied in network graphs was understood as descriptions of the context in which participants are situated, while network graphs were visual representations of the context. The textual data (interview transcripts) obtained by the first author incorporated participants’ meaning-making (Tracy, 2020) of the context and entities embedded in the context.

As shown in Figure 2, qualitative structural analysis hinges on quantitative social network analysis and reflexive thematic analysis due to its ability to “connect the relational information from the (quantitative or qualitative) network data” (Froehlich, 2020, p. 133). Because quantitative indicators are embedded in realities (Bazeley, 2018) and are essentials that depict network structures (Edwards, 2010), quantitative indicators informed the guidelines used when applying qualitative structural analysis to graphical data. The first author employed qualitative structural analysis to enrich the meanings of quantitative indicators. Since qualitative structural analysis reveals network features (e.g., key social actors) and “informed questions about their content and meaning” (Ryan & D’Angelo, 2018, p. 152), the first author paid attention to information related to network features when performing reflexive thematic analysis. Meanwhile, the first author embraced reflexivity by reflecting on how coding textual data could expand and elucidate the results obtained in the first two stages of analysis, keeping a reflexive diary and discussing with supervisors regularly for sense-checking. How the three analytical methods were employed will be elaborated in the following sections.

Describing Janine’s Ego-Net with Quantitative Indicators

Janine’s ego-net was reproduced using VennMaker, an actor-centred interactive network mapping software package, in preparation for presentation and analysis. Compared with the paper-based graphic tools used in Hogan et al.’s (2007) study, it was inconvenient for whiteboard users to draw lines between actors to indicate that the two actors knew each other. Thus, we reproduced Janine’s ego-net with the assistance of textual data obtained in the interview and sent it to Janine for checking; the reproduced ego-net can be seen in Figure 3.

Figure 3.

The reproduced ego-net.

Reproducing the Ego-Net

The first step in reproducing Janine’s network map was to create a template of concentric circles, naming each layer and putting the social actors (individuals and groups) nominated by Janine into corresponding layers around the ego. Symbols were used to distinguish women, men, and actors belonging to specific groups, based on information Janine provided in the interview. The last step was to draw lines between the actors to mark existing relationships within Janine’s ego-net according to the interview data. We made slight adjustments to the exact positions of four nominated actors (Mr. Zhang, Association, Former classmates, and Ms. B), so the relationships between Janine’s parents and these actors became visible.

Calculating Quantitative Indicators

Size, density, tie central tendency, and tie dispersion were the foci of our analysis of Janine’s ego-net. These quantitative indicators describe the general features of an ego-net because they are calculated using data on each social actor and each relationship within an ego-net (see Crossley et al., 2015). These quantitative indicators are also the results of simple descriptive statistics that could be calculated manually to allow clearer instructions and verification.

As a straightforward quantitative indicator (Crossley et al., 2015), the size of Janine’s ego-net (excluding Janine herself) was 13 because Janine nominated 13 social actors, including individuals and groups. Following the suggestion made by Gautier and Chevrot (2015), we calculated the density of Janine’s ego-net to explore whether Janine exchanged resources with members of specific groups. Thirty-one relationships existed in Janine’s ego-net. The ego-net encompassed 14 actors including Janine, meaning the number of possible relationships was 14(14 − 1)/2 = 91. Therefore, the density of Janine’s ego-net was 0.34 (rounded to the second decimal), and the value was approximately 0.23 when we only counted possible and actual relationships between nominated social actors. An individual with a denser ego-net might belong to a coherent group and exchange resources with group members frequently, but a denser ego-net could also mean that individuals lack the opportunity to exchange resources with out-group members (Lin, 1999). A value of 0.23 was close to the upper end of the empirical range reported for small ego-nets (e.g., Li et al., 2021). Therefore, we chose to expand our interpretation of the density value by performing qualitative structural analysis.

To calculate the tie central tendency of Janine’s ego-net, we assigned values ranging from 1 to 7 to each relationship to quantify the graphical data (Van Waes & Van Den Bossche, 2020). For example, we assigned a value of 5.5 to the relationship between Ms. Li and Janine because Ms. Li was described as somewhere between a best friend and a good friend to Janine. The mean of these values represents the mean tie central tendency; in Janine’s case, this value was 3.96 (rounded to two decimal places), which indicates that the strong and weak relationships in Janine’s ego-net might be similar in quantity.

Since we were interested in resource exchanges between Janine and actors engaged in different roles (e.g., workmates, family members, and former classmates) in various life domains (e.g., workplace, family, and former schools), the tie dispersion of Janine’s ego-net was calculated using Agresti’s index of qualitative variation (IQV; Crossley et al., 2015). Possible values of IQV fall between 0 and 1; a value close to 0 implies that individuals mainly exchange resources with actors from one life domain or of the same role (e.g., family members or former classmates). The IQV of Janine’s ego-net was approximately 0.98, which meant that the types of relationships making up Janine’s ego-net were highly diverse. We identified six types of relationships within Janine’s ego-net (i.e., kinships, schoolmate relationships, teacher-student relationships, superordinate-subordinate relationships, workmate relationships, and other relationships). Janine reported two kinships with parents and with Mr. Shen, two schoolmate relationships with former classmates and with Ms. M, one teacher-student relationship with Ms. Li, two superordinate-subordinate relationships with Ms. W and Ms. D, and three workmate relationships with counsellors, other teachers, and Ms. B. Additionally, Janine mentioned three actors to whom she was introduced by her parents or former classmates, including a local association, Mr. Zhang, and Ms. Wang. The relationships between Janine and these three actors were defined as “other role relationships” in this study. After calculating the squared proportions of each role relationship type, we calculated the tie dispersion of Janine’s ego-net following Crossley et al.’s (2015) instructions.

Analysing Janine’s Network Graph with Qualitative Structural Analysis

Qualitative structural analysis then allowed us to enrich our understanding of the meanings of the above quantitative indicators and to further examine the features of Janine’s ego-net. There is no predetermined sequence for this analysis (Herz et al., 2015; Herz & Altissimo, 2021); however, we recommend describing the structural features of the entire ego-net and then shifting attention to features of specific relationships and actors’ attributes.

Entire Ego-Net Features

Observations about the structural features of Janine’s ego-net have the potential to help us interpret two quantitative indicators (i.e., size and density). For example, we identified three interconnected clusters of social actors in Janine’s ego-net: family members, actors connected to Janine’s high school, and workmates in the workplace. The first two groups were situated in the inner circles and were closely related to one another. This was partly because Janine’s high school classroom teacher (Ms. Li) maintained relationships with Janine and Janine’s former classmates, and also had some interaction with Janine’s parents. Similarly, Ms. B also assisted in bridging the structural hole in Janine’s ego-net by connecting Janine’s family members with her workmates. These connections via Ms. Li and Ms. B created more relationships between clusters and thus densified Janine’s ego-net.

Features of Actors and Relationships

The identification of actors who assisted in bridging different clusters represents a transition of focus from the structural features of the entire ego-net to the features of actors, enabling further examination of actors’ attributes (e.g., their roles) and their relationships with other social actors (Herz et al., 2015). Janine’s parents were connected to six nominated actors and contributed to the expansion of Janine’s ego-net. For example, Janine’s parents introduced her to the chairperson of a local association, ultimately helping develop the relationship between Janine and the association. Janine’s former classmates were at the centre of the cluster formed by actors related to Janine’s high school. This was because Janine’s former classmates connected with Ms. Wang and Ms. Li. A teacher in another school, Ms. Wang was introduced to the cluster as the girlfriend of one of Janine’s former classmates. Ms. Li maintained a relationship with her former students, partly because she was frequently invited by Janine’s former classmates to participate in organised activities. By understanding such connections between social actors with different roles in the focal participant’s life, researchers can identify social actors who might have greater influence on the structure of the ego-net.

Incorporating actors’ attributes and the relationships between them into the analysis paved the way for the third step of qualitative structural analysis, which focussed on the features of relationships within social networks. We focussed on the multiplexity of role relationships (Herz et al., 2015), aiming to explain the tie central tendency and tie dispersion of Janine’s ego-net. Although the central tendency value implied that Janine nominated an approximately equal number of close and weak relationships, most actors who developed workmate relationships and superordinate-subordinate relationships with Janine were situated in the outer circles and defined as weak relationships, while actors related to Janine’s high school years (e.g., Ms. Li, Ms. Wang, and former classmates) were in the inner circles. In other words, close relationships and weak relationships in Janine’s ego-net were similar in quantity but differed in role relationships, an observation that was disguised by the quantitative indicator of tie central tendency. Additionally, the multiplexity of role relationships could not be directly observed in the tie dispersion value. Ms. Li was not only Janine’s former teacher and an active participant in activities organised by Janine’s former classmates, but also a relative of a close friend of Janine’s parents. Ms. B was Janine’s workmate, but she was also the only workmate whom Janine defined as a good friend. As such, the qualitative structural analysis enriched the meanings of quantitative indicators by directing the researchers’ attention to relevant details.

Integrating Results of Prior Analyses into Reflexive Thematic Analysis

The quantitative indicators and qualitative structural analysis complemented each other, describing the features of an entire ego-net and directing our attention to key social actors. To understand the resource exchanges within Janine’s ego-net and the influences of these activities, we further explored how key actors influenced Janine’s career and why. Reflexive thematic analysis was used to investigate Janine’s interpretations of (1) the content of interactions and exchanges with clusters and actors and (2) the influence of interactions and exchanges with nominated actors on Janine’s career.

The first author consulted with her supervisors, aiming to gain richer interpretations of the data rather than a consensus on the analysis (Braun & Clarke, 2021). The analysis was partly deductive because we used a theoretically informed analytical lens that focussed on the exchanges of resources with actors, and we acknowledged how prior analyses had shaped our knowledge of the phenomenon of interest. The inductive features of our analysis enabled us to understand Janine’s interactions and exchanges with actors that did not receive much attention in our prior analyses. Since the purpose of this article is to describe how the results of prior analyses can be integrated into subsequent analyses, we highlight how integration happened at the first step of reflexive thematic analysis (i.e., familiarising ourselves with the data).

Familiarisation

The first step of reflexive thematic analysis is to immerse oneself in the data set, to maintain criticality by distancing oneself from the data and questioning it, and to take notes (Braun & Clarke, 2021). By “devoting sufficient time to and paying close attention to the data prior to analyzing it” (Inayat et al., 2024, p. 5), researchers can acquire a holistic and profound understanding of participants’ contextualised experiences and enhance the quality of the codes produced. The first author familiarised herself with the interview transcript while transcribing the recordings manually and noted down analytical ideas by inserting comments in Microsoft Word (Braun & Clarke, 2021; Byrne, 2022).

Researchers usually report that they have noted patterns in their observations, interesting information, and their own feelings while familiarising themselves with data (e.g., Byrne, 2022; Trainor & Bundon, 2021), yet these authors do not elaborate on the impact of early thoughts on their analyses. We generated some ideas influenced by the theory on which this study was built, the results of the earlier analytical stage, and the questions for encouraging engagement and reflexivity proposed by Braun and Clarke (2021).

Figure 4 illustrates the genesis of some questions that the first author considered; bold type indicates contributions of prior analyses to the question list. The first author developed three initial research questions regarding the types of resources exchanged, the processes of exchange, and the influences of such activities. The results of qualitative structural analysis shed new light on these questions by drawing the first author’s attention to the necessity of investigating resource exchanges between Janine and key social actors (e.g., parents, former classmates, Ms. Li, and Ms. B). The third source was Braun and Clarke’s (2021) practical guide, which recommends that researchers think about how and why participants make meanings in specific ways and maintain reflexivity by reflecting on their own understandings and assumptions. We acknowledged that we had paid much attention to key social actors under the influence of prior analyses, but we also assumed that Janine’s workmates and superordinates might still influence her career considering the ongoing resource exchanges in Janine’s workplace.

Figure 4.

Development of familiarisation questions for reflexive thematic analysis.

Develop Codes and Candidate Themes

After reading the transcripts twice, the first author began assigning codes to the interview data by inserting comments. In the first round of coding, the first author coded not only data that had the potential to answer the research questions directly, but also information that might imply Janine’s attitudes (e.g., Janine’s comments on her parents’ expectations of her career and private life) and descriptions of Janine’s workplace (e.g., the interactions among workmates in an office). The first author then simplified and refined these inductive codes in the second round of coding (Trainor & Bundon, 2021). Since the prior analyses had helped the first author identify some key actors, she decided to pay more attention to the resources that Janine had obtained in interactions with these key actors, and conducted another two rounds of coding with this focus in mind. She also coded information related to resource exchanges between Janine and other nominated social actors. During the coding process, the first author referred to the field notes created after the interview and the analytical notes written during the familiarisation stage, hoping to capture some interesting information that she had ignored. She kept an Excel spreadsheet to document the codes that she had created in each round of coding and the relevant extracts, as suggested by Byrne (2022).

The candidate themes reported below are more semantic descriptions than latent interpretations, although reflexive thematic analysis incorporates a certain level of interpretation (Braun & Clarke, 2021). Since the analysis of Janine’s ego-net directed the first author’s attention to key social actors related to two clusters (i.e., Janine’s family and former school), the first author developed the candidate theme –We are like a family– to describe resource exchanges between Janine and those actors with whom she had developed close relationships. These social actors tended to exchange intangible resources with Janine by sharing information, providing companionship, and introducing Janine to more social actors who could provide the same things. Although they could not help Janine with her work directly, Janine noted that these interactions confirmed her belief that working in a public school could be tedious, and her description suggested that their company might have contributed to her sense of work-life balance. These social actors may thus have influenced Janine’s career plan, which was to work in international schools rather than finding a more stable job in a public school.

The second candidate theme –They are very nice people– described the interactions between Janine and social actors she placed in more distant positions in her ego-net. Janine and these social actors exchanged work-related resources, most of which were ideas about course designs and office routines. Although Janine commented that her superordinates and most of her workmates were pleasant to work with, these social actors do not seem to have greatly influenced Janine’s career plan. After considering her job prospects in the education sector and her personal preferences, Janine considered the possibility of job hopping to other international schools and developing skills to undertake part-time jobs. To summarise, regardless of the frequent and pleasant resource exchange experiences Janine had with workmates and superordinates, social actors who developed closer relationships with her appeared to have more influence on her career plan. Figure 5 presents an overview of the results obtained using three analytical methods.

Figure 5.

An overview of the results obtained using each analytical method.

Discussion

We integrated three interdependent analytical methods while analysing an interview transcript and an egocentric network graph. Froehlich (2020) points out that qualitative social network analysis presents limited methodological options to examine network features; we address this limitation by integrating quantitative indicators that capture the general features of an ego-net with qualitative structural analysis that examines and interprets specific network features. We then develop the results of prior analyses into an input for subsequent reflexive thematic analysis, shifting our attention from network features to nuances indirectly related to networks. This analytical procedure may contribute to qualitative social network analysis and broader qualitative inquiry, while also offering practical implications for educational researchers.

First of all, we build on Herz et al.’s (2015) qualitative structural analysis by expanding the scope of ego-net data analysis. Contrary to Herz et al.’s (2015) view, we understand quantitative data embedded in graphical data as easy-to-store statistical descriptions of the context (co-)constructed by participants (and researchers). Therefore, researchers can adopt established quantitative social network analysis methods to describe the context in which participants are situated. We also connect qualitative structural analysis to reflexive thematic analysis so that qualitative social network researchers are equipped with analytical tools to shift their attention from network features to network meanings. Accordingly, qualitative social network researchers might employ data elicitation tools (e.g., concentric circles) to answer theory-driven research questions that are indirectly related to social networks. This study has exemplified integrative analysis methods for three types of data obtained via interviews on online platforms, implying that future social network researchers may consider conducting interviews on any online platform with whiteboard features, which negates the need for researchers to instruct participants in using VennMaker (e.g., von der Lippe & Gamper, 2017).

This article also contributes to a broader field of qualitative research. We have demonstrated how to transparently unify quantities and qualities in data analysis (Proudfoot, 2023). Incorporating data analysis methods for numbers, graphs, and texts, our analytical practices could be adopted to enhance the quality of analysis because the data analysis procedures involved are “conducted in a precise, consistent, and exhaustive manner through recording, systematizing, and disclosing the methods of analysis with enough detail” (Nowell et al., 2017, p. 1). In our case, interpretations of quantitative indicators influenced subsequent qualitative analysis by generating meaningful knowledge about social network features, with this influence manifested as an input at the initial stage of reflexive thematic analysis. This type of input may be more welcomed by researchers who embrace a constructivist perspective (e.g., Tuval-Mashiach, 2017), since these researchers consider it central to unpack the rationale behind methodological decisions and reflect on how their prior knowledge and assumptions shape each phase of qualitative analysis. Integration in data analysis can make this reflexive process more visible and transparent than conducting separate analyses.

Based on the above discussion, we offer several implications for educational researchers who seek to integrate different analytical methods to examine socially constructed and relationally conditioned experiences:

1. Educational researchers can collect relational and contextual data more intentionally to foreground participants’ social positioning and relational experiences. Social relationships and interactions play an important role in shaping teachers’ and students’ experiences (e.g., Coppe, 2024; Witt et al., 2024). In this regard, the demonstrated use of concentric circles in this paper enables researchers to prioritise the systematic collection of contextual and relational data that captures how participants are socially situated within educational settings in future studies.

2. Educational researchers can report their strategic integration of analyses across data types to enhance methodological transparency. To this end, future research can further develop and refine analytical processes that explicitly integrate quantitative indicators (e.g., counts, network metrics) as inputs into qualitative analysis rather than treating them as parallel or separate components. Specifically, researchers are encouraged to document in detail how outputs from analyses of numerical or graphical data inform interpretive decisions at different stages of analysis, thereby strengthening analytical transparency and reflexivity.

3. Educational researchers can examine how the analytical methods proposed in this article support the study of diverse relational phenomena across research contexts and populations. Integrating social network analysis with reflexive thematic analysis enables researchers to explore how relational structures intersect with processes, norms, and lived experiences, and to assess how social conditions shape experiences and meaning-making in educational research.

Conclusion

This article illustrates how the integration of analytical methods supports constructivist and reflexive approaches by making researchers’ assumptions, decisions, and interpretive trajectories more visible. Nonetheless, the example presented in this article is not without limitations. First, this article reports integration in the analysis of data from a single case to exemplify the analytical procedure. Future researchers may consider enhancing integration in analysis by highlighting how prior results inform subsequent qualitative analysis when analysing data from multiple cases; quantitative indicators may help generate more analytical ideas when cross-case comparisons are feasible. Although the empirical findings of this methodological paper are not intended to achieve statistical generalisability due to the limited sample size, the study functions as a worked example for educational researchers who collect ego-net data in interviews and seek to move from descriptions of network features to an understanding of participants’ narratives. Additionally, we did not use more powerful software packages (e.g., UCINET) to reproduce the visual graph and calculate additional quantitative indicators. Instead, we calculated quantitative indicators that described the general features of an ego-net and then maximised the potential of qualitative structural analysis as a complement to quantitative social network analysis, which enabled us to examine specific network features and interpret quantitative indicators. Considering this, researchers may apply other advanced software packages to describe networks with additional quantitative indicators, or draw on the sequence of our analytical strategies when examining interactions between actors in social contexts, including educational settings.