- This topic has 21 replies, 17 voices, and was last updated 7 years, 7 months ago by
.
-
Posts
-
I am conducting a grounded theory study on happiness utilizing Strauss & Corbin’s approach as my foundation. However I
have not found any information on the initial sample size that would be appropriate.
I understand that due to the nature of the study there are no strict criteria for sample size
(Patton, 1990). Also that the sample size is determined by
when one achieve saturation of information. However I
assume that there should be an initial acceptable sample size. Does any one have any
information on what the range may be?In my readings of grounded theory textbooks, I’ve found recommendations ranging from 10 to 30 interviews. There seems to be no consensus; it all depends on which book you read.
try saturation… i think in qualitative .. how much ‘quality of the info/details/richness’ takes the leads over how many. 30 is far too many…i would probably forgo quantity to ensure quality.. the target is richness not census.
Your sample size is complete when there is nothing left to code in your data. No more interviews are necessary because you’ve reached “saturation”; the data does not offer any new insights on the categories, or the relationships between the categories (theoretical coding). Suspicions would be raised if you reached saturation after four interviews, but if you’re interviewing a rare occupational group (e.g., crime scene analysts in homicide cases), then you might tentatively arrive at some generalizations about their “main concerns” after interviewing 10 people.
Thank you for the feedback! Sample saturation is what I keep coming across but me dissertation committee wants a specific number to start with and that is were I have not found a definitive answer in any of the research books.
Some dissertations committees are not well versed on GT methods. You may have to defend your selection of cases on the logic of GT methods. I highly recommend any books by Barney Glaser who split off with Strauss over a dispute over the original “discovery of grounded theory”. In my humble opinion, Glaser’s use of GT is more faithful to the constant comparative method than Strauss and Corbin.
It’s not a big issues on sample size in qualitative reserach. It’s just about saturation and th richness of data. But as far as i concern 10 -30 depth interviews with help you to come up with reliable modeling in GT. It’s a hot debate regarding of Barney, Strauss and Cobin ,Charmaz and other who claimed better GT method approach. As a PhD student we have to justify whtaever we do.
You are right, there are no strict criteria, for example a very specific phenomenon could only be experienced by 5 people theoretically. In most studies you could use 30 participants as a rough (I emphasise rough) guide. In simple terms the number of participants is driven by the need for theoretical saturation – only you can determine this as concepts emerge from your data.
Here is one of (only) three evidence-based papers that I know on the topic of sample sizes in qualitative research. Hope it’s helpful.
Cheers,
Greg
Beware ! It is easy to talk glibly about ‘saturation’, but it is rarely clear when saturation is reached. A state of affairs where no new codes seem necessary may betoken that nothing more can be learned from the data, but equally it may be that the analyst simply lacks imagination to go further, or that the sample has been too small to capture the richness of the population of cases sampled. In terms of sample size, this is a classic case of the problem of induction. If, pace Popper, your first 30 cases were all white swans you might declare saturation. But if the 31st was a black swan then much of the coding established so far on white swans alone would have to be revised.
Note also that you and those commenting on your query assume the unit of study is an individual – you are asking how many individuals would make up an adequate sample. In practice however, for much of the classic grounded theory work the unit has been a behaviour or a situation – for example in Awareness of Dying it is not individuals who are the units, but responses to terminal illness, and insofar as individual patients comment on the behaviours of different doctors and nurses, one patient may be a sample of several.
I’d add another issue to consider- how much data can you handle? Imagine the pile of notes/files of recordings or whatever form your data tames in its rawest state. Now think about what you need to do. Will you have to type every word up? Will you then go through every word and code it? Even though you are going to ground your theory in the data, you probably have some ides about potential categories. How may do you think there will be at a minimum?
Are you aiming to do any ‘; theoretical’ sampling (e.g. hoping to get views from men and women to see if they differ.) If so, you might factor this in.
If you are working alone, unfunded, with short timelines relying on coding using colour highlighting or some other ‘by hand’ technique and thinking that your concepts will probably require quite complex coding (lots of categories, sub categories and sub sub categories) you might only be able to deal with ten interviewees/ observations. If you have funds, knowledge about a computer program that can help manage material for analysis (examples are NVivo and Atlas-ti), plenty of time, simple categories in mind, you could take 30 as a lower limit.
Think too about future issues. Do you plan to make a career out of this research? Then go for as much data as you can get, select some of it for current purposes and keep the rest for later articles etc.
To satisfy a committee, select a sensible number and explain why you have chosen it. (EG “I will aim to interview fifteen young and fifteen young women”), My experience has been that 30 interviewees is as many as a doctoral researcher can handle without help from other people a CAQDAS, and that practical issues like time may mean you do not quite reach the desired size.
- The forum ‘Default Forum’ is closed to new topics and replies.
