Environmental Engineering Reference
In-Depth Information
There are many textbooks available on designing studies in the social sciences
(e.g. de Vaus 2002, which focuses on quantitative surveys, or Patton 1990, which
focuses on qualitative research).
3.2.3.3 Bias reduction
The interpretation of quantitative research results requires an understanding of
two factors: the precision of the results and their bias (Section 2.2). A precise study
will have low sampling error, and so have tight confidence intervals around the
best estimate of the parameter value. An unbiased result will give parameter
estimates that are not consistently higher or lower than the true value. So it is possible
to have a very accurate estimate that is at the same time biased, or an inaccurate
but unbiased estimate (see Figure 2.1). Generally, using statistically valid survey
design methods, particularly randomisation, will guard against bias. Hence although
confidence intervals may be very wide if there is a lot of sampling error in the study,
the estimate of the mean is unbiased.
Bias is an insidious problem because it is very hard to quantify, and hence it is
difficult to correct for. However, it is pervasive in the kind of studies we are dis-
cussing, even after randomisation. For example, studies that rely on recall are often
biased by people's differing perceptions of the past and the present (people often
feel that the past was better than the present). Similarly, if you ask hunters about
their typical catch rates, you are likely to get inflated estimates, because people tend
to discount days on which they catch nothing. There will also be biases introduced
through your relationship with the interviewee; their perceptions of your motiv-
ations and allegiances and yours of theirs. There are several methods that can be
used to reduce bias, or at least check for it. These include:
●
Minimising the time-recall period (for example, ask about yesterday not last
month, Figure 3.2).
●
Asking about actual values rather than typical values (for example, ask about
numbers caught yesterday, not what is usually caught in a day).
●
Triangulating (for example, ask hunters about who they gave meat to, and then
ask consumers from whom they received it; ask focus groups about the major
points of food scarcity in the year, and then ask individuals the same question).
●
Designing the study carefully (Section 3.2.4), and using trained local research
assistants.
●
Spending a significant amount of time in the community, so that your and
your interviewees' understanding of each other's motivations is more closely
aligned.
3.2.3.4 Some golden rules
Firstly, consider carefully in advance the
practicalities and realities
of fieldwork,
particularly when you are working in a country and culture other than your own.
This includes being fully prepared for all the logistical challenges, bureaucratic
hold-ups and health and safety issues that may arise. This is not just for your own
