Environmental Engineering Reference
In-Depth Information
interview
, in which the main questions are prepared in advance and asked at some
point, but the conversation is allowed to move naturally through these questions,
so that other relevant information is obtained.
The selection of a set of people to answer the questionnaire can be made in a stat-
istically valid way (using random sampling), or by using
purposive sampling
methods (in which particular individuals or households are deliberately selected).
Fully purposive samples cannot be analysed statistically, but might be appropriate
if there are only a few people with the characteristic of interest, so random
sampling is likely to miss them. Snowball sampling is another form of purposive
sampling, for situations in which it is hard to get into the community of interest.
The researcher interviews one person and asks to be introduced to another, and so
moves through a network of contacts. In many cases, it is useful to combine
random and purposive sampling methods. For example, if the aim was to discover
whether non-timber forest product gatherers are poorer than other members of the
community, the potential respondents might first be stratified—purposely broken
down into NTFP gatherers and others—and then random households picked
from each group. It is best if at all possible to carry out some form of random
sampling, so that results can be generalised away from the sampled group through
statistical inference.
There are many textbooks on how to design a proper survey (e.g. de Vaus 2002).
There are also a number of websites where surveys are stored, so that researchers
can get ideas about how best to phrase questions (see the list at the end of the
chapter). In general, the more formal the questionnaire, the easier it is to analyse,
but the more prone the results are to researchers' preconceptions. Closed questions
(those that give respondents the option to choose between a set of predefined
answers) are particularly dangerous, and need thorough piloting. Open questions
that lead on to discussion are the best way of getting an understanding of the topic,
but are often difficult to put into a statistical framework. Here are a few pointers
and pitfalls:
●
Keep it
short and simple
. The questionnaire should not take more than 20 min to
administer and should be an enjoyable experience for the respondent (Figure 3.3).
Pilot it thoroughly, and use focus groups and key informant interviews to ensure
that you have an understanding of which questions to ask and how. Have an intro-
ductory session explaining who you are, and why you are doing the questionnaire,
and that answers are entirely confidential. At the end, thank the respondent for
their involvement and tell them how they can get more information (for example
ask if they would like a copy of the results, and give your contact details).
●
Start with
general questions
to get the session flowing, and put
sensitive questions
at the end. Ask sensitive questions in acceptable ways (for example getting people to
point to the income band that they belong to is more acceptable than asking them
directly what their income is). Think hard about the
order
in which you ask ques-
tions, so that people don't take cues from previous questions about what you are
interested in (for example if you ask people about their views on the benefits of a
