Qualitative Sampling

Qualitative Sampling

In qualitative research, sampling can occur at several stages, both while collecting data and while interpreting and reporting on it.

Sampling while collecting data for qualitative research is not the same as sampling in quantitative research because researchers are not interested in being able to generalise at a statistical level – instead the key is purposive or strategic sampling. Many would therefore argue that probability-based sampling (e.g. random sampling) is inappropriate. Sampling strategies can be determined in advance and/or evolve during the research process (gradual definition of sample structure).

Sampling Strategy Determined in Advance

Researchers might determine their strategy in advance by devising a sampling frame which has to be filled (this means drawing up a chart which lists the types of characteristics they wish their sample to include, such as a range of age-groups, a range of health status groups, a range of areas in terms of affluence/deprivation; then ensuring that recruitment continues until there is at least one “tick” against each characteristic). This is sometimes known as criterion/quota sampling because the researchers identify criteria that are important to study and select cases that meet these criteria. Another strategy determined in advance would be “complete collection”: for instance a researcher who wanted to learn about what happened to middle aged men who had chronic renal failure in an area of SE Britain would try to recruit all patients who were male, aged 30-50, in all renal units in that particular area.

Gradual Definition of Sample Structure

There are many ways of gradually defining the sampling structure, some are listed below but you may come across others. It is important to remember that practical constraints are a particular issue in many qualitative research studies (because data collection is time-consuming for example): in practice you may find that many samples are “convenience samples”. The important message to take away from this section is one about how to recognise good practice in sampling when you read about qualitative research, and how to make sure your own research is of high quality - document your sampling strategy, be able to describe the characteristics of those involved in the study, demonstrate that the sampling strategy contributes to a search for disconfirming evidence (i.e. show that you are not just collecting data that you hope will support a pre-conceived idea), and give a full and honest account of what was done. This applies to choosing illustrative quotations when reporting and disseminating research, as well as to sampling participants or events.

Theoretical sampling: here, sampling is directed towards making possible generalisations about theoretical propositions and new cases are selected for their potential in extending or testing emergent understanding.

Convenience sampling: here researchers select those who are available and likely to participate, perhaps over a specific period.

Typical case sampling: researchers decide on the characteristics of “typical” individuals and select the sample to fit the characteristics of typical cases in selecting people for focus groups etc.

Critical case sampling: researchers choose cases that they believe to be especially important because of the position they hold (e.g. because of their particular place within an organisation) or because they are especially well able to articulate a view (e.g. spokesperson for a medical charity).

Maximum variation sampling: here researchers identify cases with different characteristics to maximise diversity in the sample, usually to get the widest range of views possible. This is best done using quota methods (see “strategy determined in advance”) and iteratively (e.g. refine your quotas based on the emergent analysis – you may decide you need to investigate one group in more depth). Any common themes that emerge can capture “core” experiences or views.

Intensity sampling: this is where researchers sample the same characteristics over and over. It tends to be used in interpretive phenomenological analysis.

Snowball sampling: relies on referrals, one participant recruits others. This can help researchers to capitalise on informal networks that might otherwise be difficult to access (e.g. sex workers, drug users, victims of domestic violence etc).

How do you Know When you Have an Adequate Sample?

Ideally this will occur when your quota (sampling frame) is full, and when you have reached theoretical saturation – in other words when additional cases do not modify your coding frame. It is a good idea to have some redundancy, and to be able to report on when you reached saturation and how. (Theoretical saturation occurs when new data from new cases do not contribute to the development of emerging theory even after you have tried to ensure that your new cases are those most likely to extend or challenge your ideas.)  In practice, if you are applying for funding for a study, you will need to specify how many participants you are likely to need. As a rough guide, for interview studies analysed using constant comparative approaches, theoretical saturation will probably be reached after 20-60 interviews.

Samples for Focus Group Studies

There are some additional points that might help when arranging focus group studies. The recommended size of a group is 6 – 10 people, having fewer than this could limit the potential interaction, and having more than this could make it difficult for everyone to join in the discussion. A study using focus groups to collect data should include several groups, not just one, because any individual group is subject to internal or external factors of which the investigator may be unaware. Also, on a practical level, there may be individual groups that do not go smoothly: the members may be reluctant to participate or not interact well with each other and limited insight will be gained. Enough groups should be run to provide adequate breadth and depth of information; there is a tendency for 10 to 15 groups per study but some studies may find that 4 or 5 are enough. There is no upper limit on the number of focus groups that could be held although this will be limited by resources.

The members of each focus group should usually have something in common: characteristics which are important to the topic of investigation. For example, they may all be members of the same profession or they may work in the same team. They may all be patients at a practice or have experienced a similar health problem or be receiving similar treatment. Participants might or might not know each other. There are advantages and disadvantages to both.