Environmental Engineering Reference
In-Depth Information
needs to understand the material in order to draw correct inferences. For example,
Whitley (1994) discusses the socio-cultural reasons for the rock art he studied
being full of representations of hunters killing bighorn sheep, despite the fact that
the society was predominately seed-eating.
As might be expected from its intellectual tradition, there is a wide range of methods
for analysing
qualitative data
. These include arts-based analysis in which the
researcher collects evidence, thinks about its meaning and then constructs an inter-
pretation based on their experience and understanding. This may be aided by the
construction of diagrams illustrating linkages, flows and overlaps between actors
and processes.
Much qualitative research is based on detailed
textual analysis
of field notes
and interview transcripts (Strauss and Corbin 1990; Coffey
et al
. 1996; Seidel
1998). This highlights patterns in the data, which can then be analysed system-
atically. There are a number of software packages available that can be used to
automate the process (see links at the end of the chapter). The software allows
the researcher to search efficiently for particular words or phrases. It can then
identify juxtapositions between these words, and hence discover patterns. So
for example, a detailed textual analysis of a particular hunter interview may sug-
gest that whenever hunter X is discussing his motivations for hunting he talks
about his cultural identity. Then a broader search of all the interview transcripts
could be used to confirm that people do indeed seem to link hunting to their
cultural identity.
Quantitative data are analysed using
statistical techniques
. However, the analysis
should always start with simple data exploration. This includes graphing the data so
that a visual impression of relationships is obtained. Then simple univariate analyses
can be carried out, such as Chi-squared tests, correlation analyses or t-tests. More
complex statistics to analyse the effects of multiple factors and their interactions can
then be used, based on the understanding obtained in the exploratory phase of data
analysis (Chapter 4). These include general linear models, logistic regression and
correspondence analysis. Many statistics textbooks are available, online or in print
(see Sections 3.4 and 4.6 for suggestions).
Model-based analyses
are also useful tools. In this case, the data are used to para-
meterise a model, which expresses how the researcher understands the system. For
example, a model of hunter behaviour might test the hypothesis that hunters
actively conserve their resources by ignoring hunting opportunities in depleted
areas (Alvard 1993). The parameterised model is then validated against independent
data to show how closely it predicts reality. Rowcliffe
et al
. (2003) used data on prey
densities, prey encounter rates with snares and probability of snares catching the
prey in a model to predict offtake rates. They could then validate their model using
actual offtake rates from the same systems. We explore these kinds of models in
more detail in Chapter 5.
