Grounded theory analysis
I have one questionnaire on which the data is collected from consumers and organisations level, now the data that is collected from both respondents around 25, i want to know when i am going to analyse the data collected by grounded theory Straussian approach, is that means 2 grounded theory will be emerge or can i analyse date from both of them( consumers and organisations level)
Views: 324
Replies to This Discussion
-
Permalink Reply by Pat Bazeley on -
Given you have used the same questionnaire, your initial analyses will be with both groups together (to get descriptive information, at least), but then you will compare your two main groups, consumers and suppliers, for each question or issue raised. Whether you can build a common theory will depend on how different their views are, e.g., if they use the same dimensions or constructs in answering questions (even if they are at different poles on those) then you will have a common theory. I'm not sure that I'd call your work Straussian grounded theory, however, if you have used a "questionnaire' (that sounds like it was very structured, where GT would call for a more open approach)?
-
Permalink Reply by Mahmoud Hassan on -
Hi Pat,
it was typing mistake, because i did not use questionnaire i have used interview question that spread among two type of participants consumer and organisational level by adopting strussian approach in analysis method , so is that need to have two grounded theory or just only one theory can emerge from both two type of participant. as far i know just use their response separately in open coding and then integrate those categories which emerged from open coding in the axial coding by using paradigm model and then develope theory. so my question related to is that need to analyse date from each type of participant or as i mentioned above i mean one theory will be emerge
-
Permalink Reply by Roger Gomm on
-
In constructing your questionnaire, you presupposed what was important and what was relevant for your research. In any grounded theory approach, relevance, importance, and theoretical meaning are supposed to emerge from the analysis of data. In most excerises trying to derive grounded theory, further data collection is supposed to follow on from and be guided by the analysis of the first tranche of data - and the research may then repeat this several times until a statisfactory theory emerges. So unless you are going to use the results of your questionnaire just as a device to collect exploratory data on which to base further research (and not questionnaire research), there is no way your approach can derive a theory which is grounded in the data, rather than a theory structured by the ideas you had which led to you ask the questions you featured in constructing your questionnaire.
I take it then, that what you mean when you say you are going to use a grounded theory approach is that you are going to code the data along the lines suggested by Strauss and Corbin. If so, then given the remarks above, you would be advised to say you are going to use Straussian coding, rather than claim to be discovering theory grounded in the data. The most likely result of using Straussian coding is that you will rediscover the features you structured into the data by using a questionnaire in the first place, although you will probably also discover how inadequate your questionnaire was - from a grounded theory point of view all questionnaires are deficient.. Avoiding the premature imposition of a structure on data (as occurs in using questionnaires) is precisely what Strauss and Glaser were arguing against in 'The Discovery of Grounded Theory' and what has been carried forward in other works by Strauss, Corbin, and with some differences, by Glaser.
-
Permalink Reply by Mahmoud Hassan on
-
Hi Roger ,
it was typing mistake, because i did not use questionnaire i have used interview question that spread among two type of participants consumer and organisational level by adopting strussian approach in analysis method , so is that need to have two grounded theory or just only one theory can emerge from both two type of participant. as far i know just use their response separately in open coding and then integrate those categories which emerged from open coding in the axial coding by using paradigm model and then develope theory. so my question related to is that need to analyse date from each type of participant or as i mentioned above i mean one theory will be emerge
-
Permalink Reply by Roger Gomm on
-
Well I'm not quite sure what you are doing, but the 'grounded theory' answer to your question of whether you will find/create one theory or two (or 3, 4, 5 etc) is that the answer will be found by analysing the data: that's what 'gounded' means. However, since you asked similar questions of two different sorts of respondents, it is almost invevitable that your coding will feature similarities and differences in response between the two kinds: you built that into the data by asking the same questions of two different groups. In that sense you have already shaped the data up to be most suitable for generating a single theory explaining why there are similarities - where there are similarities, and why there are differences where there are differences in response - or at least any other approach to the data would look rather perverse.
-
Permalink Reply by Ahmad Salih on -
Hi Mahmoud,
What do you mean by 'consumer' and 'organizational' level? Can you provide more details? Are you sure you used a questionnaire? Questionnaires are normally used in quantitative studies!
Is the total of your sample 25 participants? Are you trying to compare data collected from both levels? Or will you aggregate the two data sets in one pool?
In call cases, according to the guidelines of grounded theory, you are supposed to start data analysis as soon as you collect your first set of data (first interview or two, first focus group, first observation, etc.). Have you done that?
-
Permalink Reply by Mahmoud Hassan on
-
Hi Ahmed ,
it was typing mistake, because i mean question not questionnaire and i have used interview question that spread among two type of participants consumer and organisational level by adopting strussian approach in analysis method , so is that need to have two grounded theory or just only one theory can emerge from both two type of participant. as far i know just use their response separately in open coding and then integrate those categories which emerged from open coding in the axial coding by using paradigm model and then develope theory. so my question related to is that need to analyse date from each type of participant or as i mentioned above i mean one theory will be emerge
-
Permalink Reply by Laurence Moseley on
-
There are two problems to be solved before you can draw any conclusions:
1. The sample - is it representative of anything wider than itself? If it is not, then you should draw no conclusions. Even if your measurement, checking of your interpretation, recording of the raw data, coding framework, and allocation of codes to text items were done as accurately as possible (with many checks to weed out the possibility that you, or some other data analyst, was biased) if the sample is not representative all you can do is to describe it - not draw wider conclusions.
2. Measurement and interpretation - are you confident that the recorded textual material is accurate (e.g. not biased by any leading questions or by order effects). Recall that Schuman and Presser found that, using an identically questionnaire but reversing the order of two of the identically worded questions, they produced estimates of support for abortion as different as 18% and 51%. How you obtained the free-text utterances of your respondents is a possible source of bias and you need to be able to rule out such bias.
Are you confident that no biases have crept into your coding? Was the coding done by two or more coders independently (i.e. not knowing what the other coders had done)? That is what I have done in the past. Can you interrogate your data to see how much support there was for the main themes that you extracted. For example, were some mentioned by more people than were others? Were some coded as expressing stronger view than were others? To give one quotation to illustrate a particular theme has no information value unless it supported by some indication of the strength or frequency of occurrence of the theme.
Do not worry about criticism for using a questionnaire. I have frequently used highly structured questionnaires which allow some questions to be open-ended. For example, in a study of midwifery-led care, in addition to the structured questions and scales, there were two free-text ones : "What was the best thing about your pregnancy?" and "What was the worst thing about your pregnancy?" They still produced free-text data which were highly informative.
Have you thought what your unit of analysis should be (i.e. how big is each chunk of text which you will code)? In another study, the coding was done by two people independently. One of the many factors being coded was the emotional tone of a statement (with the values Positive, Negative, Mixed, or Neutral). On the first run-through, after the coding had been completed the level of agreement between the two coders was calculated. It was only 45%. Clearly there was something wrong with our coding. Such an inter-coder comparison should always be carried out.
On examination, it turned out that there were many comparative statements (e.g. "My previous pregnancy was terrible, but the current one was wonderful"). One coder had treated that as one textual item and had coded it as "Mixed". The other coder had treated the two pregnancies as different text items and had coded one of them as "Negative" and the other as "Positive". That was the reason for our lack of consensus.
There were two positive results of this episode. The first was that we handed the coding framework to a third person. She did not know how we had coded each item, only that the first two coders had disagreed. The first two coders agreed on the coding of 87% of items (much better consensus), and with the third coder included that rose to 97%. The remaining 3% we threw away. After all, if three people cannot agree what an utterance means, it is reasonable to treat it as meaningless.
The other benefit was that we learned something about how to divide up a piece of text into meaningful chunks. That makes the coding more precise and avoids large chunks of text which are classified as "Mixed" or some other uninformative label.
One does not need to define "qualitative" as "non-quantitative". It may be less structured, but you can still count aspects of the process. Most people would give greater weight to a view uttered by many people than to one uttered by only one person, even if that one person happens to agree with the researcher's prejudices.
HTH
Laurie
-
Permalink Reply by Roger Gomm on
-
'if three people cannot agree what an utterance means, it is reasonable to treat it as meaningless'
Laurie,I strongly agree with your comments on the benefits of inter-coder checks on coding, except for the phrase above which equates meaninglessness with disagreements. It may be that where coders disagree, researchers can't do anything with the utterances concerned, as is also the case usually where a single researcher can't understand what a respondent says. But uselessness isn't the same as meaninglessness, and it is often where people disagree that they invest the most meaning in their own version. It seems to me that inter-coder checking shouldn't just be used as an editing device to delete items of disagreement, but should extend to drawing attention precisely to those utterances where alternative interpretations are possible. This is, after all, a common feature of everyday speech.
Of course, you were not really writing about meaningfulness in the sense of the meanings subjects might attribute to their own utterances, but classifiability in terms of the coding schemes devised for analysis of the data; hence it would have been better to say -'reasonable to treat it as unclassifiable in terms of our coding scheme'. A different coding scheme might have elicted more (or less) agreement between coders - as you illustrate.
Nonetheless, it simply is a feature of human language use that utterances are sometimes formulated to evade classification, and are often ambiguous as to what type of utterance they are: take for example the frequent disputes as to whether an utterance should be taken as an insult/ a joke/ a mistake (and note the temporal dimension in this example,whereby what might have been intended as an insult, is thereafter agreed to be a joke or a mistake or vice versa). Disagreements here between coders wouldn't render the utterance meaningless and ought to classify such an utterance as one of the types of utterance subject to everyday disagreement over its classification.
In my experience utterances which have led to coder disagreements, and some which I have found most puzzling myself, have turned out to be the most theoretically interesting.The problem , of course, is deciding which might be worth pursuing further and which should be thrown into the unclassifiable bin.
-
Permalink Reply by Laurence Moseley on
-
Roger - I too agree. I did write "it is reasonable to treat it as meaningless", not "it is meaningless". The quotation is about what you do when you get disagreements.
However, without a great deal of further evidence (probably external to the actual text) one should not merely assume that it is meaningful. I have done a lot of work on meaning in questionnaires, using a variety of mechanisms to try to assess whether a given utterance (free text in qualitative work, or a question in a survey questionnaire, or a stub in a scale) can be treated as meaningful. In one study I threw away 91% if the proposed questions on the grounds that we could not achieve consensus about their meaning among a group of judges who had been selected to be representative of theultimate respondents
There's a whole discussion to be had on the meaning of the word "meaning", However, language is a tool for social communication, and therefore trying to assess whether the initiator of an utterance and the recipients of it all interpret it in the same way is a reasonable way of proceeding. Do you agree?
-
Permalink Reply by Roger Gomm on
-
Laurie
I think much depends on your initial decisions as to what the data are to stand for. In questionnaire research for example, it is usually assumed that 'good' questions, elicit responses from which some knowledge, attitude, belief or such like of respondents can be inferred accurately - there are good and bad questions in this regard: good questions elicit responses which are both accurate (as above) and comprehensible: bad questions elicit reponses which are either inaccurate and/or incomprehensible, or elicit no response at all. And in this regard testing the effeciency of questionnaires as you suggest is crucial.
By contrast, when using naturally occuring speech as data one can treat the utterrances as data about the uses people are making of language in this particular interaction. Then silences, ambiguities, humourings,understatements, exaggerations and so on are not there to be repaired (as might be the case in questionnaire research), so as to 'see behind' them to see true feelings, beliefs,attitudes and such like.It remains useful though to conduct some inter-coder checking, but the questions for coders are different. Not ' what did he mean by that ?, perhaps, 'but did this utterance by a second party treat the utterance by the first party as (say) a seriously held opinion/ a consensual or co-social response/ a put down ?' (it being subsequent utterances which usually serve to establish what earlier utterances mean for that moment- and of course the same utterance can change its meaning in the course of interaction, and can be put to different uses by different parties).
-
Permalink Reply by Laurence Moseley on
-
I agree. What is apparently one analysis may in fact be several analyses (some reading the body language, some the silences, some the textual utterances, etc). Before we can start to put all those different meanings together we have to be confident that individually they are accurate and pose (and preferably answer) some relevant research question. That is one of the reasons why the analysis of the use of natural language is so difficult and requires such methodological convolutions to prevent any individual passing off their own prejudices as some sort of evidence. When one is dealing with communication (in whatever form) which is essentially a social activity involving at least two people, the measurement of consensus between recipients or observers about the meaning of what is being observed becomes useful.
It does not matter what the utterer intended. If the recipient of the message (words, frown, smile, shudder, whatever) understands it in a different sense from that intended then communication has not occurred. We need to know what the recipient understood by the communication. That is why I find it so much more useful to question people about what they (or the utterer) would do rather than what they think. I was an adviser on one study which used (despite my protests) a 5-point Likert-type scale to assess the amount of support for various health-care policies.
The lowest point was "Strongly Disagree" - a not uncommon choice. However, I asked what might people who had answered "Strongly Disagree" actually do. Within about 2 minutes I had come up with at least 11 courses of action which different people might take. Those courses of action ranged from "I would resign my post rather than implement that policy" through "I would have to arrange retraining for my staff" to "I would feel dissatisfied if it was implement, but would grind my teeth and suffer in silence". In that case we might have done better to ask quite simply "Would you resign?"
The general lesson for beginners is that designing, executing, analyzing, and reporting research is not a matter for amateurs. My own usual estimate is that you should spend about 25-30% of your research time on devising those steps i.e. in design and development. I am still shocked when a new research student turns up and wants to start by going out and interviewing people - with no idea of the choices that he or she has to make before gathering their first datum. Whoever trained them must have thought that research was simply a well-paid version of gossip!
One thing that is rarely noted is that market research company and opinion survey organisations often spend some time on trying to delve into something called "meaning". However, at the end of the day they usually get paid for predicting which goods people will buy or which candidates the public are likely to vote for. That is another example of what you do is the important element of meaning.
© 2013 Created by SAGE Publications.
