|RESEARCH SERIES NO.19
|Year : 2019 | Volume
| Issue : 1 | Page : 40-45
Data analysis in qualitative research
College of Nursing, CMC, Vellore, Tamil Nadu, India
|Date of Web Publication||09-Oct-2019|
Dr. Vinitha Ravindran
College of Nursing, CMC, Vellore, Tamil Nadu
Source of Support: None, Conflict of Interest: None
Data analysis in qualitative research is an iterative and complex process. The focus of analysis is to bring out tacit meanings that people attach to their actions and responses related to a phenomenon. Although qualitative data analysis softwares are available, the researcher is the primary instrument who attempts to bring out these meanings by a deep engagement with the data and the individuals who share their stories. Although different approaches are suggested in different qualitative methods, the basic steps of content analysis that includes preparing the data, reading and reflection, coding, categorising and developing themes are integral to all approaches. The analysis process moves the researcher from describing the phenomenon to conceptualisation and abstraction of themes without losing the voice of the participants which are represented by the findings.
Keywords: Categories, codes, data analysis, qualitative research, theme
|How to cite this article:|
Ravindran V. Data analysis in qualitative research. Indian J Cont Nsg Edn 2019;20:40-5
| Introduction|| |
Qualitative data analysis appears simple to those who have limited knowledge of qualitative research approach, but for the seasoned qualitative researcher, it is one of the most difficult tasks. According to Thorn, it is the complex and elusive part of the qualitative research process. Many challenges that are inherent in the research approach makes the analysis process demanding. The first challenge is to convert the data from visual or auditory recording to textual data. As qualitative approach includes data generation through sharing of experiences, it becomes fundamentally necessary to record data rather than writing down accounts as the stories are shared. Essential data which may become apparent or uncovered when reflecting on audiotaped interviews may be missed or overlooked if interviews are not recorded. Although field notes are written they often augment the experiences conveyed by participants rather than being the primary data source. Therefore, the researcher needs to spend effort and time to record as well as transcribe data to texts which can be analysed.
The second challenge is managing the quantum of textual data. One hour of interview may produce 20–40 pages of text. Even with fewer participants as generally is in qualitative research, the researcher may have many pages of data which need coding and analysing. Although software packages such as NVivo and Atlas-ti are available, they only help to organise, sort and categorise data and will not give meaning to the text. The researcher has to read, reflect, compare and analyse data. The categories and themes have to be brought forth by the researcher. The third challenge is doing data generation and data analysis at the same time. Concurrent data generation and analysis is a predominant feature in qualitative research. An iterative or cyclic method of data collection and analysis is emphasised in qualitative approach. What it means is that as the researcher collects data, the analysis process is also initiated. The researcher does not wait to complete collecting data and then do analysis. Iterative process enhances the researcher to focus on emerging concepts and categories in subsequent interviews and observations. It enables the researcher to address the gaps in the data and get information to saturate the gaps in subsequent contacts with earlier or new research participants. Sufficient time and resources are needed for sustaining the iterative process throughout the research process.
The above challenges are mentioned at the beginning of this article not to discourage the researchers but to emphasise the complexity of data analysis which has to be seriously considered by all researchers who are interested in doing qualitative research. In addition to the general challenges, data analysis in qualitative research also varies between different approaches and designs. There is also the possibility of flexibility and fluidity that enhances the researcher to choose different approaches to analysis, either one specific approach or a combination of approaches. The framework for the analysis should, however, be made explicit at the beginning of the analysis. In qualitative research, the researcher is a bricoleur (weaver of stories) who is creating a bricolage.
| Characteristics of Data Analysis|| |
In qualitative data analysis
- Researcher attempts to understand the meaning behind actions and behaviours of participants
- Researcher becomes the instrument to generate data and ask analytical questions
- Emphasis is given to quality and depth of narration about a phenomenon rather than the number of study participants
- The context and a holistic view of the participants' experience are stressed
- The research is sensitive to what the influence he/she has on the interpretation of data
- Analytical themes are projected as findings rather than quantified variables.
Process of data analysis
Qualitative data analysis can be both deductive and inductive. The deductive process, in which there is an attempt to establish causal relationships, is although associated with quantitative research, can be applied also in qualitative research as a deductive explanatory process or deductive category application. When the researcher's interest is on specific aspects of the phenomenon, and the research question is focused and not general, a deductive approach to analysis may be used. For example, in a study done by Manoranjitham et al., focus group discussions were conducted to identify perceptions of suicide in terms of causes, methods of attempting suicide, impact of suicide and availability of support as perceived by family and community members and health-care professionals. Focused questions were asked to elicit information on what people thought about the above aspects of suicide. The answers from participants in focus groups were coded under each question, which was considered as categories and the number of responders and the responses were elaborated under the said questions as perceived by the participants. Deductive process in qualitative data analysis allows the researcher to be at a descriptive level where the results are closer to participants accounts, rather than moving to a more interpretive or conceptual level. This process is often used when qualitative research is used as a part of the mixed methods approach or as a part of an elaborate research study.
In contrast, the inductive process which is the hallmark of qualitative data analysis involves asking questions of the in-depth and vast data that have been generated from different sources regarding a phenomenon., The inductive process is applicable to all qualitative research in which the research question has been more explorative and overarching in terms of understanding the phenomenon in peoples' lives. For example, in Rempel et al.'s study on parenting children with life-threatening congenital heart disease, the researchers explored the process of parenting children with a lethal heart condition. Volumes of data generated through individual interviews with parents and grandparents were inductively analysed to understand the 'facets of parenting' children with heart disease. Inductive analysis motivates and enhances researchers to rise above describing what the participants say about their experience to interpretive conceptualisation and abstraction. The process of deduction and induction in qualitative data analysis is depicted in [Figure 1].
| General Steps in Data Analysis|| |
Although different analytical processes are proposed by different researchers, there are generally four basic steps to qualitative data analysis. These steps are similar to what is generally known as qualitative content analysis., In any qualitative approach, the analysis starts with the steps of content analysis. The content analysis ends generally at an interpretive descriptive level. Further analysis to raise data to abstraction may be needed in some approaches such as grounded theory.
- Preparation of data
- Reading and reflecting
- Coding, categorising and memoing
- Developing themes/conceptual models or theory.
Preparation of data
As already discussed, the inductive process in qualitative research begins when data collection starts. Each recorded data set from individual interviews, focus groups or conversations should be first transcribed and edited. The researcher may decide on units of data that can be analysed to further help in organising. The units can be the whole interview from one individual or interview transcripts from one family or data from different individuals connected with in a case (as in case study). On some occasions, the unit may consist of all answers to one question or one aspect of the phenomenon. Many researchers may not form any such units at the beginning of the analysis which is also accepted. The essential aspect of the preparation is to ensure that participants' accounts are truly represented in transcribing. Researchers who have a large amount of content will need assistance in transcription. One hour of interview may take 4–6 h to transcribe. An official transcriber will do a good job than a researcher who may spend a long time in transcribing volumes of data. However, the researcher has to edit the transcription by listening to the audiotaped version and include words and connotations that are missed to maintain accuracy in transcription. Another important point to note is to transcribe and prepare the data as soon as interviews are completed. This facilitates the iterative process of data collection and analysis. All data, including field notes, should be organised with date, time and identification number or pseudonym for easy retrieval. Assigning numbers or pseudonyms help to maintain the confidentiality of the participants.
Reading and reflecting
Reading the data as a whole, and reflecting on what the participants are sharing gives an initial understanding of the narrative. The reflection may start at the time of the interview itself. However, reading and rereading the transcribed text from an interview gives an understanding of context, situations, events and actions related to the phenomenon of interest before the data can be analysed for concepts and themes. Reading and reflection help the researcher to get immersed in the data, understand the perspectives of participants and decide on an analytical framework for further data analysis. As texts are read, the researcher may jot down points or questions that are striking or unusual or does or does not support assumptions. Such reflective notes assist the researcher to decide on questions to be asked in further interviews or look for similarities or differences in interview texts from other participants. These initial reflections do not complete analysis; rather, it provides a platform for the analysis to develop. An example of initial reflections when analysing interviews from a study on home care of children with chronic illness is given below.
Reflections-family 1 interview
'This family has a lot of issues related to home care. Their conversation is a list of complaints about the system and the personnel. Even though it appears that help is being rendered for support of child at home, nothing seems to satisfy the parents. The conversation revolves around how they have not been given their due in terms of material and personnel support rather than about their sick child or the siblings.
After a while, it became tedious for me to read this transcript as I resent the complaints (which I should not do I suppose). I wonder how other families perceive home care.'
The initial reflections also help to understand our position as a researcher and the assumptions the researcher brings to the study. It helps us to be aware of one's own professional and personal prejudices which may influence the interpretation of data.
Coding, categorising and memoing
For analysis to progress further the researcher has to decide on an organised way of sorting and categorising data to come to an understanding about the phenomenon or the concepts embedded in the phenomenon. Researchers may choose to analyse only the manifest content in a descriptive qualitative study or may move further to look for latent content in an analytical-qualitative study. The manifest content analysis includes looking for specific words or phrases used by the participants and accounting for how many have expressed the same or similar words/phrases in the data. It looks at what is obvious. Latent content analysis, on the other hand, involves coding and categorising to identify patterns and themes that are implicit in the data.
Coding is an essential first step in sorting and organising data. Codes are labels given to phrases, expressions, behaviours, images and sentences as the researcher goes through the data. It can be 'in vivo' codes or 'interpretive codes'. When participants' exact expressions itself are used as codes it is called 'in vivo' codes. If the researcher interprets the expression or behaviour of the participant depicted in the text, then it is called interpretive codes. In the grounded theory method, different levels of coding are suggested. The first level is called the 'open coding' that involves sifting through the initial data line by line and creatingin vivo or interpretive codes. Questions such as what are this person saying or doing or what is happening here? will help in the initial coding of data. Initial coding may reveal gaps in the data or raise questions., These gaps and questions will help the researcher to locate the sources from where further data are to be collected. The second level is known as 'focus or selective coding' will be used in subsequent interviews. Focused coding involves using the most frequent or most significant earlier codes to sift through large amounts of data. Focused codes are more directed, selective and conceptual and are employed to raise the sorting of data to an analytical level. The first level of coding can be done manually or can be done using qualitative software packages. In other types of content analysis, the different levels of coding may not be followed instead the researcher engages in interpretive coding as the text is read. In a grounded theory study on parenting children with burn injury open codes such as scolded, accused, unwanted, guilt, nonsupport, difficult to care, terrible pain, blaming oneself and tired came up as the data were coded [Table 1]. These codes gave the researcher an initial insight into the traumatic experiences that the parents undergo when caring for their burn-injured children. As texts were coded, the researcher attempted to understand further the struggles of parents in the successive interviews with other families.
|Table 1: Selected codes from the analysis of transcripts on parenting children who had sustained burns|
Click here to view
Categorising involves grouping similar codes together and formulating an understandable set within which related data can be clubbed. A category is 'a collection of similar data sorted into the same place' – the 'what', developed using content analysis and developing trajectories and relationships over time. It is a group of content that shares commonality. Data can be categorised generally when the researcher realises that the same codes or codes that are relatively similar are emerging from the data. When categories are developed based on codes, they can be still at descriptive level or can be at an abstract level. By developing categories a conceptual coding structure can be formulated. At this level, there is no need to continue line by line coding. Instead, the researcher uses the coding structure to sort data. In other words, parts of data that best fit the categories, and the codes are grouped appropriately from across the data sets. The grouping of data into categories is enabled by comparing and contrasting data from different sources or individuals. As constant comparison continues questions such as 'What is different between the accounts of two families? What are similar? Will help in grouping data into categories. As the researcher compares data, questions such as 'what if' may come up which will propel the researcher to return to participants to know more or even purposively include participants who will answer the question. The data under each category should be read again to ensure that they appropriately represent the category. Qualitative software packages are very useful in sorting and organising data from this level. Any part of data which is not fitting into any category needs to be coded newly, and the new codes should be added. The emerging new codes may later fit into a category or form new categories. All data are thus accounted for during this phase of analysis.
As analysis and grouping of further data continue, the researcher may rearrange data within categories or come up with subcategories. The researcher may also go from data to codes, to sub-categories which then can be abstracted into categories. In the burn study, similar codes that were repeated in many transcripts were grouped together. Grouping these codes helped in developing subcategories such as physical trauma, emotional trauma, self-blame and shame. The sub-categories were then grouped to develop meaningful categories such as facing blame and enduring the burn [Table 2]. Creating categories thus assists the researcher to move from describing phenomenon to interpretation and abstraction.
|Table 2: Selected codes, categories and themes from analysis of transcripts on parenting children who had sustained burns|
Click here to view
Memoing is 'the researcher's record of analysis, thoughts, interpretations, questions and directions for further data collection' (pp 110). Memos are elaborations of thoughts on data, codes and categories that are written down. Simply put, memoing is writing down the reflections and ideas that arise from the data as data analysis progresses. As data are coded, the researcher writes down his/her thoughts on the codes and their relationships as they occur. Memo-writing is an on-going process, and memos lead to abstraction and theorising for the write-up of ideas. Initial or early memos help in exploring and filling out the initial qualitative codes. It helps the researcher to understand what the participants are saying or doing and what is happening. Advanced memos help in the emergence of categories and identify the beliefs and assumptions that support the categories. Memos also help in looking at the categories and the data from different vantage points. One of the early memos from burn study is given as an example.
24 June, 2010, 10 pm – After coding interview texts from three families.
'I am struck by the enormity of a burn injury. I realize that family members cannot do many things for the child at home after discharge of a severely burned child because the injury is so big that even some clinics and doctors who are not familiar with burn care cannot manage care. These children need continuous attention of the health care professionals. They need professional assistance with dressing. They need professional assistance with splints and gadgets, and therapies. The injury is extensive that it is difficult for family members to do many things on their own. It is very hard, very hard for the parents to take up a role of the caregiver for children with burns because it involves large wound which has not healed or is in the process of early healing and the child suffers severe pain. The post burn care is very different from caring for other children with chronic illness or congenital defects which most often does not involve pain. The child's suffering makes it easy for the parents to view them as vulnerable. Yet the parents do their best. They try to follow the Health Care Professionals advice, they try to go for follow-up, but it seems simply not enough. I think the parents are doing all that they can within the context of severe injury, lack of finances, lack of resources in home town, or blame and ridicule from neighbors and others…'
Stopping to memo helps the researcher to reflect on data, move towards developing themes and models and lay the ground for discussion of findings later. Memos need to include the time, date, place and context at which they were written.
Developing themes, conceptual models and theory
Developing themes involves the 'threading together of the underlying meaning' that run through all the categories. It is the interpretation of the latent content in the texts. Theming involves integrating all the categories and explicating the relationship in the categories. In coding and categorising the researcher is involved in deconstructing or dividing the data to understand the feelings, behaviours and actions. In the phase of theming, the researcher is trying to connect the deconstructed part by understanding the implicit meaning that connects the behaviour, actions and reactions related to a phenomenon. To identify theme, the grounded theorist asks: What is the core issue which the participants are dealing with? The phenomenologist will ask about the central essence or structure of the lived experience related to the phenomenon of interest. The ethnographer may look at the cultural themes that link the categories. The researcher generally comes up with one to three themes. Too many categories or themes may indicate that the analysis is prematurely closed and implies the need for the researcher to further interpret and conceptualise the data. In the study on parenting children with burn injury, the researcher came up with the theme of 'Double Trauma' which explicated the experiences of parents living the burn with their children and also enduring the blame within the context of both the hospital and home [Table 2].
In phenomenology and ethnography, the analysis may end with identifying themes. In other approaches, such as grounded theory and interpretive description, the analysis may progress further to developing theory or conceptual models. Identifying the core category/variable from the coding activity, memos and constant comparisons are the first step in moving towards theory development in grounded theory. The core category is the main theme that the researcher identifies in the data. The next step in grounded theory is to identify the basic social process (BSP). The BSP evolves from understanding how participants are dealing with the core issue. In real-world situations, individuals develop their own strategies and process to deal with the core issue in any situation. Identifying this process is the stepping stone to theory development in grounded theory. In the example of burn study, the theme 'Double Trauma' was the core category and parenting in the burn study involved a dual process of 'embracing the survival' and 'enduring the blame'., A conceptual model was developed based on these processes.
| Pitfalls in Qualitative Analysis|| |
Large data sets for analysis
As already explained, the amount of data text or field notes from observations and other sources in qualitative research can become overwhelming if data analysis is not initiated concurrently with data collection/generation. Coding large data text is tedious and takes much of the researcher's time. Postponing analysis to the end of data collection also prevents the researcher from becoming focused in subsequent interviews and filling gaps in data in further data collection. Therefore, deferring data analysis should be avoided.
Researcher should not hasten to conclude analysis with developing categories or themes. This may lead to 'premature closure' of the research and the danger that the participants' experiences are misunderstood or incompletely understood. Qualitative data analysis involves in-depth interaction with the data and understanding the nuances in the experiences and the meanings behind actions. The researcher continues to generate data until all the categories are saturated, which means that the categories are mutually exclusive and can be explained from all aspects or angle. In the burn study, although the table in this article appears simple, the codes and categories were developed from larger data sets representing multiple participant interviews and field notes. The category 'facing blame' was brought forth with parents' accounts of experiencing blame in almost all the families in one or multiple ways: from family members, health-care professionals, strangers and the child itself. The researcher needs to be reflexive and iteratively do data generation and analysis until there is no new information forthcoming in the data. Inferring conclusions too soon which is otherwise known as 'inferential leaps', will prevent the researcher from getting the whole picture of the phenomenon.
Interpretation of meanings
During the analysis process as the researcher interprets and conceptualises the participants' experiences, he/she delves into the tacit meanings of actions and feelings expressed by participants or observed in various situations. The researcher endeavours to keep the interpretations as close to the participants' accounts as possible. However, it should be understood that the meanings are co-constructed by both the participant and researcher by collaborative effort which is also a hallmark of qualitative research. In the process of co-construction, researcher should be cautious to not lose the voice of the participants. Discussion with peer at all steps of analysis or checks on codes and categories by others in the research team may help to avoid this problem.
| Conclusion|| |
Qualitative data analysis is a complex process that demands much of reading, thinking and reflection on the part of researcher. It is time-consuming as the researcher has to be constantly engaged with the texts to tease out the hidden meanings. Beyond the differences in data analysis in different qualitative methods, coding, categorising and developing themes are the essential phases of data analysis in most methods. Researchers should avoid premature conclusions and ensure that the findings are comprehensively represented by participants' accounts. Qualitative data analysis is an iterative process.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Thorne S. Data analysis in qualitative research. Evid Based Nurs 2000;3:68-70.
Holloway I, Galvin K. Qualitative Research in Nursing and Healthcare. England: John Wiley and Sons; 2016.
Bringer JD, Johnston LH, Brackenridge CH. Maximizing transparency in a doctoral thesis: The complexities of writing about the use of QSR* NVIVO within a grounded theory study. Qual Res 2004;4:247-65.
Mayan MJ. Essentials of Qualitative Inquiry. California: Left Coast Press; 2009.
Lincoln YS, Denzin NK, editors. Turning Points in Qualitative Research: Tying Knots in a Handkerchief. California: Rowman Altamira; 2003.
Philipp M. Qualitative content analysis. Forum Qual Soc Res 2000;1:10.
Manoranjitham S, Charles H, Saravanan B, Jayakaran R, Abraham S, Jacob KS. Perceptions about suicide: A qualitative study from Southern India. Natl Med J India 2007;20:176-9.
Rempel GR, Rogers LG, Ravindran V, Magill-Evans J. Facets of parenting a child with hypoplastic left heart syndrome. Nurs Res Pract 2012;2012:714178.
Polit DF, Beck CT. Essentials of Nursing Research: Appraising Evidence for Nursing Practice. New Delhi: Lippincott Williams and Wilkins; 2009.
Graneheim UH, Lundman B. Qualitative content analysis in nursing research: Concepts, procedures and measures to achieve trustworthiness. Nurse Educ Today 2004;24:105-12.
Mabuza LH, Govender I, Ogunbanjo GA, Mash B. African primary care research: Qualitative data analysis and writing results. Afr J Prim Health Care Fam Med 2014;6:E1-5.
Bradley EH, Curry LA, Devers KJ. Qualitative data analysis for health services research: Developing taxonomy, themes, and theory. Health Serv Res 2007;42:1758-72.
Morse JM, Richards L. ReadmeFirst for a User's Guide to Qualitative Methods. New Delhi: Sage publications; 2002.
Strauss AL, Corbin J. Basics of Qualitative Research: Grounded Theory Procedures and Techniques. Newbury Park, CA: Sage Publications; 1990.
Glaser BG. Theoretical Sensitivity: Advances in the Methodology of Grounded Theory. Mill Valley: CA; Sociology Press; 1978.
Glaser BG, Strauss AL. The Discovery of Grounded Theory: Strategies for Qualitative Research. Mill Valley, CA: Sociology Press; 1967.
Charmaz K. The grounded theory method: An explication and interpretation. In: Emerson RM, editor. Contemporary Field Research: A Book of Readings. Boston: Little Brown; 1983.
Morse JM. Confusing categories and themes. Qual Health Res 2008;18:727-8.
Speziale HS, Streubert HJ, Carpenter DR. Qualitative Research in Nursing: Advancing the Humanistic Imperative. USA: Lippincott Williams and Wilkins; 2011.
Strauss A, Corbin J. Basics of Qualitative Research Techniques. Thousand Oaks, CA: Sage Publications; 1998.
Charmaz K. Constructing Grounded Theory: A Practical Guide through Qualitative Analysis. Thousand Oaks: Sage; 2006.
Ravindran V, Rempel GR, Ogilvie L. Embracing survival: A grounded theory study of parenting children who have sustained burns. Burns 2013;39:589-98.
Ravindran V, Rempel GR, Ogilvie L. Parenting burn-injured children in India: A grounded theory study. Int J Nurs Stud 2013;50:786-96.
[Table 1], [Table 2]