Hello all, I'm in the process of coding semi-structured interviews and I'm a little (very!) confused as to what I'm supposed to be doing. I know the coding system goes like open coding, axial coding and selective coding.
I've analyzed my texts, read them numerous occasions, but what exactly are the "themes" I'm supposed to be finding? I've tried line by line and added numbers to what *I* think are pertinent pheonema in the text. But on what basis is one supposed to decide what a code is? Are they at my discrection or should I be using a theoretical framework? Where does Grounded Theory come into this?! Should I have commenced the research with a theory in mind and attempted to "test it out" as it were?
I've attempted to code the interviews and sent it to my lecturer but I feel like I've over generalized and made some very gross assumptions with the data. I also noticed, re-reading the coding exercise that I've erred on the side of quantitative in that I've stated how many times the respondents answered a certain question in a particular way. Is that incorret? I'm very new to all this so I hope this doesn't come as across as a particularly elementary problem!
Any help would be greatly appreciated!
Just to add to Dave's comments, you have to come out with the themes by yourself. The themes can be developed based on the reviewed of the previous studies you did on your study together with the objectives/research questions of your study. From these you will notice from your transcription and coding that your informants gave views which relates to the objectives/the research question. Themes emerge based on the emphasis informants/participants lay on a particular issue or phenomenon and how issues raised by informants relates to the objectives/research questions(the thesis).
On whether you've erred on the side quantitative, it is important to know these two basic things. In qualitative research we use participants or informants instead of respondents. Again, you erred because if you for instance stated that 'many respondents', 'respondents mentioned 4 times', 'majority stated that' etc. Instead, in qualitative research, you will have to replace these (quantitative assertions) with qualitative statements like: 'predominant views from informants/participants', 'there were recurring views/statements/comments from informants/participants', 'informants/participants frequently/repeatedly commented' etc. These make your work more qualitative rather stating that 'many respondents' or 'how many time'....because the moment you state 'many' then it gives opportunity to people who will critique your work to ask how 'many' of the respondents did x or y? And as you may be acutely aware, this type of question will make you give figures hence making you work more of quantitative (statistical figures) rather than qualitative (words).
Remember your work has to be original and the originality of your work will depend on the new ideas you capture from your informants which actually forms the themes of your study for better analysis and discussion as well conclusion.
I hope this will be helpful.
I am very willing to help... is it ok sir if you will send me the copy of the interview in text so that I can evaluate your process and suggest for possible theme(s) for that. I taught research methodology both descriptive and qualitative here in the Philippines for quite a long time. I will wait for your reply. Thank you.
I am facing dificulty in anaylsing the semi-strucured interview can I ask about what is the wy to analyse it, do I have to used coding or I can do it normal way.
Please refer Sage publication's book - Coding manual for qualitative researchers by Johny Saldana.
Refer Basics of qualitative research by Corbin or Strauss. I not exactly remember but plz find on net. It shows axial coding technique with example.
Thanks so much for all the replies everyone, I'm still trying to figure it all out! Slow progress!
It sounds to me like what you are actually doing is content analysis, which is more of a quantitative method than a qualitative.
As you didn't make your decision prior to conducting your interviews you are limited which method of analysis you can apply to the data. You can't do IPA because the interviewing technique should be phenomenological, not standard, if you are doing grounded theory then you need to collect data and analyse simultaneously to ensure theoretical saturation, this is not the same as data saturation for thematic analysis.
Thematic analysis is a flexible method and if you choose to do that then you shoudl read the article by Braun and Clark. That will help you with the coding and subsequent analysis. This article will also tell you how to make sense of the themes that emerge from the data
remember that there are some quality criteria for qualitative methods and you need to think about these when you make your choices. Most qualitative methods are theoretically driven and have certain procedures for conducting them properly so you can't just decide half way through to use them
in general, searching for the themes is done by reading the transcripts first and getting familiar to it. then the inquirer looks for the salient ideals to group them under big concepts which are called themes.
To answer your question, “what basis is one supposed to decide a code is?”
Actually, what you intend to code depends on what you consider the data is relevant and important to your research. In the middle ground, there is no standard method of coding in qualitative analysis but as long as you follow the appropriate procedure of coding, then it will be fine. Codes may be descriptive, merely labelling the salient aspect of the text. In regard of developing codes, you have to assign a descriptive word or phrase to each unit of notes. Here are a few steps for starting points in developing codes:
To answer your question, “Where does Grounded Theory come into this?”
Importantly, the intent of a grounded theory is to generate a theory from field data. It is the context of a phenomenon being studied that provides the basis for a grounded theory study. The sample of individuals for a grounded study is selected on their ability to contribute to the development of the theory. Often, a homogeneous sample is selected first, one in which each individual has had a similar experience. Once the theory is developed, a heterogeneous sample, individual who have had different experiences may be selected to confirm or disconfirm tenets of the theory. Research problems in grounded theory studies are focus on what happen to individuals, why they believed it happen as it did, and what it means to them. Furthermore, the questions for grounded theory are broad and general and is usually followed by sub questions related to coding of the data. The researcher needs to learn as much as possible from the participants via interview about the participants’ perceptions of the nature and impact of the experiences as related to a possible theory.
The systematic process of grounded theory through data analysis are as below:
Form the initial categories and subcategories from the information,
To answer your question, “I've over generalized and made some very gross assumptions with the data, is that incorrect?”
Here are some tips for you. If pursuing in doing generalizations:
I hope those information can assist you in doing your qualitative research.
Estimado Austin, disculpa que responda en español, pero el inglés no es un idioma en el que soy fuerte.
En principio, la técnica que estás utilizando es la apropiada. Usualmente, solemos partir de una estructura conceptual que nos dirige el análisis (entiéndase, un conjunto de categorías y subcategorías de análisis que pretenden descomponer y explicar el problema de estudio en el que te estás centrando. Luego de que has elegido un conjunto de categorías de análisis, puedes ir creando nuevas categorías, a partir del momento en el que las frases no "encajan" en tus categorías iniciales.
Un aspecto importante que debes considerar es que cada palabra, cada verbo, cada adjetivo debe guardar correspondencia con la categoría de análisis previamente determinada. Es cierto que a veces una frase en su conjunto denota más semánticamente que sólo palabras inconexas. Por esta razón, se recomienda que -luego de que realices tu primera codificación en categorías- le des el producto de tu trabajo a una tercera persona (preferible más de una persona) conocedora del tema, para que ésta evalúe la consistencia interna de la clasificación que has realizado. Es el criterio de jueces externos quien termina ayudándote a "afinar" la clasificación inicial y el análisis posterior.
Espero que te haya sido de utilidad. Si encuentro algún texto que te pueda ayudar, te lo enviaré. Pero creo que estás bien prientado en el tema, por lo que expones.
I get a lot of good stuff from the following:
Creswell, Dr. John W. (2008-07-15). Research Design: Qualitative, Quantitative, and Mixed Methods Approaches (Kindle Locations 575-576). Sage Publications, Inc. Kindle Edition.
Recall that, even when accurately recorded and in context, a selected verbatim quotation merely tells your reader that one person said one thing on one occasion. How you interpret that is your choice. If you alone do the coding, all that the reader has is your opinion, and can reasonably be sceptical.
If the verbatim quotation is given as an example of a point made by some other means then it may be more acceptable. It is also useful to explain to your readers what you meant when you gave that code to a particular utterance.
The other way to avoid the potential criticism that all that you have written is merely your opinion is to have the coding done by two or more coders. Determine a coding framework. Using that, code each utterance yourself. Ask a colleague to use the same coding framework to code each utterance independently of you i.e. they may have a list of which codes you use, but not to which utterances you allocated them.
You can then do a formal comparison of the two sets of coding. If codes agree, you treat them as accurate. If they do not, then you do not allocate any code at all to that disagreed utterance. In one study in which the coding was done independently by myself and another coder, after the first round of coding we agreed on only 40% of utterances. Where we disagreed, a third coder recoded the disagreed items. Note that the third coder did not know what codes the first two had allocated - merely that they had disagreed.
You have to be careful to define what you mean by an utterance. In fact, part of the skill is to divide up your text into meaningful chunks. In the example above the first two coders agreed on only 40% of the utterances. When we looked to see why, it was largely a question of time. When a respondent had said "My last pregnancy was horrible but this one was wonderful” I had coded that as two utterances - one negative (the horrible one) and one positive (the wonderful one). The second coder had coded the whole chunk as "mixed". That was one of the reasons for calling in a third coder. So it mattered whether an utterance was defined as "a statement about one experience" or a "statement about several experience, albeit in the same sentence or clause". Spend some time on that definition.
Once I have read the text I determine the variables which I wish to code. For example, you might have variables like
Who was it about e.g. doctor, nurse, midwife, administrator....
Where was it about e.g. hospital, surgery, clinic, the respondent's home....
What was it about e.g. competence of staff member, personality of staff member, information giving, feeling of control...
What was its tone e.g. positive, negative, neutral, mixed, not relevant
Was it before or after some clinical intervention
... and many other things
You can then see whether there is consensus among the coders on each of the utterances and each of the variables.
To permit that to be done I have used a relational database. I used Paradox, but it can be done in Access, MySQL or whatever you happen to know or can learn.
Hence, if you want to include a quotation about positive evaluations of competence of staff members you can add up the counts for doctors, midwives, etc and the verbatim quotation can then be supported by that count. No one can reasonably say "that's just your opinion" as you can "no it isn't because 3 (or however many you have used) coders independently agreed on those codes". You can say things like "68% of all comments about midwives were positive, for example 'My midwife taught me a great deal' ". In other words the verbatim quotation is in context and is supported by the fact that it was not made by only one person. That, combined with the fact that the final coding was consensually agreed, should remove much of the potential criticism.
The psychologically hard bit is when you find no consensus. In that case, strictly you should say that the utterance is meaningless and should be excluded from the analysis. Would you agree that if 3 people working independently cannot agree on whether an utterance is about a midwife, a hospital, is positive or negative then it is fair to say that it is meaningless?
The themes emerge naturally by an accumulation of the variables. If a lot of the utterances are complaining about the personalities oif foctors, midwives, and other, you may have a theme which is "dissatisfaction with the personalities of staff". You can easily build up large themes from smaller ones. It is much harder the other way round. Having built them up, you can easily retrieve the verbatim quotations which led you to do the coding as you did. that way, your quotations are always meaningful and in context.