During Week 1, we looked at the fundamentals of quantitative data analysis, including sampling, descriptive statistics, and correlational statistics. Later in the course, we will spend some time looking at inferential statistics. For the next several weeks, we will be exploring the fundamentals of qualitative data analysis, including coding, visually displaying data and analyses of data, and making a final report of the research findings. This week, our focus is on chapter 4 of the Qualitative Data Analysis (Miles et al., 2020) text.
In review, quantitative data analysis focuses on numbers and statistics; qualitative data analysis focuses on words and writing; mixed methods approaches focus on both. Regardless of the research methodology, getting started with data analysis can be daunting. As you begin the process of analyzing the data you have collected, it is wise to begin by clarifying your purpose.
In his famous book The 7 Habits of Highly Effective People (first published in 1989), Stephen Covey advises that we begin with the end in mind. For scholarly researchers, this means revisiting your research problem, purpose, and questions. One mistake that novice researchers sometimes make is to completely forget about these things by the time they report their findings or results!
As you begin the process of data analysis, consider typing your research problem, purpose, and questions in large font on a piece of paper and posting it in your work area so that you don’t lose sight of why you started conducting the study in the first place. In chapter 4 of your scholarly research project, each of your research questions should be explicitly answered (or not if you don’t have the data to answer sufficiently). Even if you don’t have the data to answer sufficiently, you will need to state this clearly in chapter 4.
As you analyze your data over a period of weeks and months, your goal is to think deeply about what you are discovering so that you can make assertions and propositions. According to the Qualitative Data Analysis text, assertions are “declarative statement[s] of summative synthesis, supported by confirming evidence from the data and revised when disconfirming evidence or discrepant cases require modification of the assertion” (p. 92). Propositions are “statement[s] that put forth a conditional event, an if-then or why-because proposal that gets closer to prediction or theory” (pp. 92).
It is never too early to begin making assertions and propositions about your research findings, but don’t settle on your final analysis prematurely. Miles and colleagues suggest that you keep a running list of assertions and propositions, review them often, and revise them as you continue to collect and analyze data. For example, assertions and propositions can be used to guide further data collection. If you have interviewed three people and initially asserted that professors in the Department of Chemistry are suffering from low morale, interview three more professors and ask questions to test your assertion.
If the new data that you collect indicates that some professors in the department are feeling fulfilled and optimistic, you may need to modify your assertion to, some professors in the Department of Chemistry are suffering from low morale. You may then be motivated to develop a proposition to explore why some professors are experiencing low morale and some are not. Hopefully, you can see how assertions and propositions can deepen and enrich your data analysis as long as you don’t draw conclusions too quickly.
Remember, also, that you need to remain mindful of research limitations during the data analysis stage of your study. Don’t draw conclusions too quickly; don’t ask leading questions; be sure that every assertion and proposition is supported with multiple examples or sources of evidence, etc. If you are careful to maintain the integrity of your study from start to finish, your research will be more credible and trustworthy. If it is credible and trustworthy, others will be more likely to read it and, more importantly, apply it. The reason you are doing all of this work is to make a difference, so make sure that you meticulously adhere to best practices throughout the process.
One strategy that I have found useful during the data analysis stage is to maintain an audit trail, or a running log of my data analysis activities. In addition to keeping myself on track and guiding my next steps, I have found the audit trail useful in completing the final report of my research methods (your chapter 3) because I can simply report what I actually did. Your ENC 706 Data Analysis Journal would be one way to get started with an audit trail for your study.
If you’ve already read chapter 4 of the Qualitative Data Analysis text, you may have noticed that I’m taking Covey’s advice and beginning with the information at the end of the chapter. Before we get into the details of qualitative coding, it makes sense to me to review the big picture.
Regarding case-based analytic approaches, Miles and colleagues distinguish between within-case analysis, a very detailed exploration of a single case, such as one person, one department, one event, etc. and cross-case analysis, which explores several cases. While within-case analysis provides “a well-grounded sense of reality” (p. 94) about one case, it is not necessarily applicable to similar cases. The advantages of cross-case analysis are the increased chances of generalizability, or application to other contexts and the opportunity for deeper understanding and explanation due to the comparison and contrast of multiple cases. The takeaway here is that multiple data sources (in within-case analysis) and multiple cases (in cross-case analysis) will render rich analyses, assertions, propositions, and ultimately your study’s final conclusions.
Of the two analytic approaches, Miles and colleagues further distinguish between a case-oriented approach, which conducts a holistic exploration of each individual case before
comparing and contrasting across all cases, and a variable-oriented approach, which focuses on specific variables and their interrelationships across all cases with little attention to individual cases. Of the two approaches, variable-oriented analysis is more generalizable than case-oriented analysis.
Miles and colleagues are of the opinion that a variable-oriented approach to cross case analysis is superior in terms of generalizability, understanding, and explanation. However, the analytic approach a researcher selects must relate directly to the research problem, purpose, and questions. Even if you are not conducting a case-based study, this information may be helpful as you revise chapter 3 of your scholarly research project.
Now that we’ve considered where we’re going, let’s return to the beginning of the chapter. Qualitative data analysis begins with coding the data you have collected. In preparation for coding, your observation notes should be typed, your interviews should be transcribed, survey comments should be downloaded and organized, etc. Any data that you have collected that is represented with words can and should be coded.
So what are codes, anyway? A code is simply a label that assigns meaning to each chunk of data in your study, which can range from a short phrase to a full sentence to several paragraphs. The researcher’s notes and questions can be coded as well. Subcodes are codes that fall under the umbrella of a larger code to further distinguish the data. For example, the code “anxiety” may include the subcodes “academic,” “financial,” and “social.”
First cycle codes are the labels that are initially assigned to each chunk of data. Once first cycle coding is complete, second cycle codes, also called pattern codes, are used to refine and reorganize the data chunks into meaningful categories that help to address the research questions.
You have several options in regard to how you manage your data during the coding and analysis stage. Some researchers prefer to do everything on paper, by literally cutting data into strips and using Post it notes to make memos as they form initial assertions and propositions. Other researchers swear by Computer Assisted Qualitative Data Analysis Software (CAQDAS), such as MAXQDA and QUIRKOS. Some versions are even free, such as QDA Miner Lite. Each product has different features, which can be very helpful in keeping you organized as you analyze. However, you may experience a steep learning curve, which you may or may not have time for.
A third, in-between option is to use Microsoft Excel to organize your data. Enter sections of data, such as individual comments, responses, sentences, paragraphs, etc. into each row. Provide identifying information for each section of data, such as the date, the participant’s name or location, etc. for easy sorting later. I also like to include three columns for codes, just in case I can’t decide on just one code. Anything that gets coded two or three times is revisited at a later time as the analysis becomes more nuanced.
What might your column headings look like in Microsoft Excel, if you were to code participants’ interview responses? You can sort by date, by participant, by interview question, or by code, depending on what you are looking for. Excel also offers the option of color coding, which can be an easy way to get a sense of which codes are most predominant. Always keep a back up of your in-progress analysis!
Again, the purpose of first cycle coding is to categorize the data collected into meaningful chunks for deeper analysis. You have a plethora of options for first cycle coding. For example, there are several elemental coding methods that provide a foundation for later coding and analysis.
Descriptive codes assign labels to data that describe or summarize the gist of the data with a word or short phrase. Descriptive codes are usually, but not always, nouns. Recall that nouns include people, places, things, and ideas. In vivo codes are exact words or phrases taken directly from participants’ responses. For example, if a participant said, “It felt like an out-of-body experience!” an in vivo code might be “out-of-body experience.” Process codes are gerunds (or -ing words) that label both observed and interpreted action in the data. Using the previous participant response, a process code could be “experiencing” or “feeling.” Concept codes represent broad, often abstract ideas such as “time” or “ambivalence.”
As a researcher, you can mix and match code types however you wish. In vivo coding might be a good place to begin. You might also prefer to simply start coding and then analyze the types of codes that are emerging as you interpret the data.
If you don’t care for the previously mentioned coding types, you’re in luck. There are more! According to Miles and colleagues, the three affective methods “tap into the more subjective experiences we encounter with our participants” (p. 67). Emotion codes label either the emotions recalled and/or experienced by participants or inferred by the researcher.
Values codes are labels that signify values, attitudes, and beliefs expressed by participants or inferred by the researcher. Miles and colleagues define values as the importance attributed to ourselves, other people, things, or ideas. Attitudes are the way we think and feel about ourselves, other people, things, or ideas. Beliefs are worldviews or schemata made up of one’s values, attitudes, personal knowledge and experiences, opinions, prejudices, morals, and other perceptions of the world. Clearly, values coding involves quite a bit of interpretation on the part of the researcher and is best confirmed by research participants, if possible, before the final report is published.
Evaluation codes are symbols, usually plus (+), neutral (x), or minus (−) that assign judgments about the merit, worth, or significance of programs or policy under study.
One literary and language coding method is dramaturgical coding. This type of coding explores human action and interaction through strategic analysis of people’s motives. Common codes for dramaturgical coding include “objectives,” “conflicts,” “tactics,” “attitudes,” “emotions,” and “subtexts.” Usually, in a second round of first cycle coding, these primary codes are supplemented with subcodes that provide greater detail. For example, “tactics” might be subcoded “friendliness” or “withholding information.”
There are also three exploratory coding methods, which are particularly useful if the researcher isn’t quite sure where to begin. Holistic coding applies a single code to a large chunk of text, a paragraph, an entire page, an entire section, to capture its overall essence. Holistic coding allows the researcher to code a data set fairly quickly before returning for a second round of first cycle coding that digs deeper into the data.
Some researchers prefer to develop a list of codes beforehand. In this case, provisional codes can be developed from the research questions or from what the researcher anticipates will be discovered. Provisional codes are only a starting point and will need to be revised, modified, deleted, or expanded as the researcher immerses in the data and begins to understand it better. One common mistake novice researchers sometimes make with provisional codes is forcing the data to fit the codes. If you choose to use provisional coding, plan on doing a second round of first cycle coding at a later time to refine your analysis.
Hypothesis coding is similar to provisional coding in that the researcher begins with a pre-determined list of codes, based on a researcher-generated hypothesis. Especially if you are conducting a mixed methods study that includes a correlational or inferential statistical analysis, hypothesis coding may be an appropriate choice.
If you are conducting a study that aims to explore or evaluate a system, you may wish to use a procedural coding method. Protocol coding tracks preestablished, recommended, standardized, or prescribed systems. For example, if you are studying college students who have recently lost a loved one, you might code interviews using the five stages of grief. If you wish to describe college students’ study habits, you might develop a coding system that includes “taking notes in class,” “assigned readings,” “homework,” “studying.”
Just as its name indicates, causation coding is used to track cause/effect sequences. The researcher develops codes that label antecedents that eventually lead to a final outcome. The plus (+) symbol is used to signify a combination of antecedents and the greater than (>) symbol signifies that one antecedent leads to another. In the previous example about college students’ study habits, a researcher might find that “taking notes in class” + “viewing demonstrations on YouTube” > “above average course grades.”
Believe it or not, there are four more types of first cycle codes from which you can select. There are four grammatical coding methods. Attribute coding labels basic descriptive information, such as participant demographics, professional title, or other variables. Attribute coding is another good place to begin, but keep in mind that some of this information can be included as a column in your Excel spreadsheet so that you have more latitude for more analytical codes later.
Magnitude codes are best used in a second round of first cycle coding, to indicate the intensity, frequency, direction, or presence of first cycle codes. Generally, symbols such as plus (+), double plus (++), zero (for neutral), minus (−), and double minus (−−) are used. Keep in mind that with magnitude coding and all qualitative codes, once you’ve made sense of the data, the next question is always, why?
We have already discussed subcodes, which are labels that supplement primary codes, such as “anxiety: financial.” If you find you have subcodes for everything you are coding, it is a sign that you may need to revise your coding scheme.
Simultaneous coding is the application of two or more different codes to a single chunk of data. For example, you may have a rich paragraph of text that describes or explicates two very different concepts. Rather than using the code/subcode strategy, it is better to maintain both concepts as primary to see if they develop further as you continue to collect and analyze data.
If there is one main takeaway from all of these first cycle coding options, it is that as the researcher, you get to decide how to code and analyze your data. There is no right or wrong way to do it. However, you will be expected to describe your analysis process in your final research report, so keep track of what you are doing as you work using an audit trail.
Let’s transition now to an overview of creating, managing, and revising qualitative codes. Qualitative coding is an iterative process that, to some degree, you will develop as you immerse within it. The open-endedness and creativity of the process is what I love the most about qualitative research!
You may have noticed that some of the first cycle coding methods we have discussed are deductive or created in advance. Researchers sometimes refer to these as a priori coding systems. Other methods are inductive or generated from the data collected. Researchers refer to inductive codes as empirically grounded within the data. Both approaches are credible, but empirically grounded codes are often viewed by others as more trustworthy or credible.
Miles and colleagues provide sage advice about coding lists and level of detail. It is very important that your codes have structure and unity. Whether you develop an a priori list of
codes or begin by coding inductively, make sure you take the time early on to list all of your codes and subcodes on a single page and look for overall categories, sequences, etc. Look at your research purpose, problem, and questions for guidance. Always remember that the reason you are coding the data is to reveal information that will help you develop thick descriptions and/or offer plausible explanations in response to your research questions. Generally, as you conduct your analysis over a period of weeks or months, you will go from many codes to fewer, more refined codes, to three to five overall categories or themes that are rich with nuanced details. Revise your codes, assertions, and propositions often as your understanding of the data deepens.
In addition to developing, and frequently revising, your coding list, it is a good idea to write definitions for your codes. The definitions will not only help you to be precise and consistent as you code each section of data; you may wish to use some of the definitions in your final research report. Again, being very clear about what you mean by “power struggle” or “professional aggression” will make your study’s findings and conclusions more trustworthy.
One final word of advice. Qualitative data is abundant, and it can be overwhelming in quantity. Moreover, however interesting, much of it will not be of value to your research purpose, problem, and questions. For example, I can think of funny things that research participants have said during interviews that I very much wanted to be able to include in my findings and discussion, but they weren’t necessary. You will need to let these things go. As a researcher, I am always reluctant to let data go too soon, in case I might need it later. So, I always create a code such as “unnecessary data” for things that I don’t think I will need. This allows me to set this data aside so that I can focus on what I perceive as the more relevant data. Later, as my understanding deepens, I go back and review everything coded as unnecessary to see if I missed something. This strategy may be useful to you as well.
You are probably beginning to understand that first cycle coding may take a while. There may be periods of time when you need to pause for several days or weeks just to think through what you’ve coded before continuing with the process. This is a good thing. Analyzing data requires thinking. Even when you are taking a break to think you are engaged in the process of data analysis.
By the time you get to second cycle coding, or pattern coding, you should know your data like the back of your hand. Through the use of refined codes, sorted data, revised assertions, and confirmed or disconfirmed propositions, you also should have a fairly clear idea of the direction your analysis is going. The purpose of second cycle coding is to condense the data you’ve coded so far into fewer categories, to test your interpretations of the meaning of the data with confirming or disconfirming evidence, and ultimately to deepen the analysis you have begun.
Pattern codes usually consist of categories or themes, causes and explanations, descriptions of relationships, and/or theoretical constructs. Often, these concepts or patterns overlap. For example, a researcher may have articulated the theoretical construct of “servant leadership” from the data and elucidated participants’ expression of servant leadership by identifying four or five categories or themes. Within those categories and themes, it is possible that causes and explanations are described or that relationships are exemplified.
By the time you get to second cycle coding, you will be nearing the end of the data analysis stage of your scholarly research project. But how do you actually analyze all of the data that you’ve coded? First, be sure to qualify or test all of your pattern codes. What may be true for some participants may not be true for all participants. One way to qualify codes is to refine your propositions, or if-then statements. For example, if survey respondents who strongly agree that they are receiving a strong liberal arts education also comment that they do not feel well versed in world languages and culture, you may need to revise your proposition to reflect this contradiction. Another strategy is to assert varying explanations. For example, these students may not have taken courses in world language and culture yet. Or perhaps the students are used to the term foreign language and may have misunderstood the question.
It can take a while before a researcher is ready to make propositions and assertions about the meaning of the data, let alone refine them. There are four general strategies that qualitative researchers use to think through the data and what the data mean. Narrative description involves writing a story that elaborates on a pattern code. Stories are structured with a beginning, middle, and end. They include a setting, characters, and a plot. Sometimes, writing a narrative description of a pattern code helps a researcher to fill in the gaps or identify weaknesses. The narrative description may be revised and later incorporated into the final research report. It can also be coded as data or used to create a visual display.
Visual displays can take the form of matrices, such as tables, or networks such as flow charts. The next five chapters of the Qualitative Data Analysis text will go into great detail about options for visually displaying data in ways that communicate its meaning. Just like listing all qualitative codes and subcodes on one page, visually displaying data can help the researcher to discover and elucidate patterns and themes that relate to the research problem, purpose, and questions.
While narrative description and visual displays are often used during second cycle coding, they can be employed at any time in the data analysis process. The same is true for jottings and analytic memos. For example, a researcher might jot down a question such as, why do students who think they are getting a strong liberal arts education also feel deficient in world languages and cultures? Such a question prompts the researcher to spend more time exploring this particular finding.
An analytic memo is a more detailed “note to self,” where the researcher reflects upon and attempts to synthesize different aspects of the data into an interrelated whole. Just like narrative descriptions, analytic memos can be used to “fill in the gaps,” identify weaknesses, and create visual displays. They also can be revised and incorporated into the final research report.
Most of the time, these strategies are used only for data analysis; the readers of scholarly research projects, dissertations, and scholarly journal articles never see them in the final report. However, sometimes they can be the crown jewel of your research. To ensure the maximum credibility of your study, you will want to support your final propositions and assertions with empirical research such as findings from similar studies. This is why chapter 4 of your scholarly research project is called Findings and Discussion.
The great thing about qualitative data analysis is that it is all up to you, as the researcher. You just need to make sure you maintain a record of what you are doing and why (an audit trail) so that you can describe it in your final report.
This brings us to the end of this presentation. Please refer to the weekly content to continue completing the Week 2 objectives and assignments.