Ray asks: "Have focus groups really become accepted as part of social
science investigation or statistical method?"
The answer to the first part of this is "yes". Social science is primarily
about systematic investigation, and focus groups are a powerful research
tool. The primary rationale for focus groups is based on phenomenological
theory: people form views, not in isolation, but through inter-subjective
social processes. Focus groups allow you to investigate the perceptions
that people share. When I teach my students about this, I usually give them
an example of shared social perceptions in action. I ask them to imagine
that a young friend is distressed about the end of an affair, and they find
her asleep on the sofa with an empty bottle of pills and a note which says,
"I want to die. Leave me alone." What do they do? In nearly every case,
they give the same answer. They have not discussed the issue, the issue is
not trivial, they have not consciously been taught or told what to think,
the answer is not the only reasonable possibility, and many of them have
never discussed the question before - but they've come, without realising
it, to think the same way. That's inter-subjectivity in action. I would
argue that, beyond marketing, focus groups are the method of choice for
discussing issues of norms, values and ethics.
The answer to the second question is "no", but it's the wrong question.
Focus groups aren't part of statistical method, and although some returns
from focus groups can be analysed statistically, the process is suspect.
Statistical method is generally probabilistic, and stochastic surveys are
governed by the general rule, "garbage in, garbage out". Focus groups often
present problems in sampling, in data production and in processing, with the
possibility of a numerical bias at each point. If the material is being
processed with stochastic processes, we might reasonably object. But the
kind of research Ray is criticising is not statistical research - it's
social research about confidence in statistics, which is completely
different. The sample described is purposive, rather than
probability-based, and the methods which are appropriate are the methods
used with purposive samples. The main criticism to make is about the
rationale of the purposive sample, not about their representativeness.
Paul Spicker
******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
*******************************************************
|