Environmental Engineering Reference
In-Depth Information
In quantitative research, there is a more concrete tradeoff to be made. This is
between the time and financial resources available to do the work and the accuracy
of the results that are obtained. A particularly useful tool for assessing this tradeoff
is power analysis (Cohen 1992). This involves estimating the probability that you
will fail to detect an effect (for example a relationship between the price of a wildlife
product and hunter effort) when it in fact exists. The power of a study depends on
the sample size, the effect size and the value chosen as the threshold for significance
(alpha,
). Alpha measures the probability that you reject your null hypothesis
when in fact it is true (i.e. detecting a relationship when there is none), and is often
set at 5%. Hence you can improve the power of your study in three ways:
●
by increasing your sample size,
●
by increasing your alpha (e.g. to 0.1 rather than 0.05), which means you will
get more false positive results,
●
by increasing the size of the effect that you wish to detect.
In general, strong relationships between variables will be picked up even if there
is a lot of noise, but weak ones will come through only with large sample sizes or if
error is reduced. This can be done by stratifying the sample into more homoge-
neous groups or including covariates. Wealth ranking can be a very useful stratifi-
cation. For example, if people of similar wealth act in similar ways (for example in
their meat consumption decisions), statistical tests that divide people into wealth
groups are likely to have more power than tests on the whole population.
Power analysis techniques range from very simple formulae to complex proced-
ures. There are a number of software packages available to help you with this (see
Section 3.4 for details).
3.2.3.2 Randomness and representativeness
This is another area in which qualitative and quantitative research differ drama-
tically. In quantitative research, the key issue is to ensure that the sample is repre-
sentative of the population as a whole. The best way to do this is to use stratified
random sampling. Stratification (into groups of similar datapoints, for example by
community, by wealth class) reduces the amount of unexplained error, and hence
increases power. Randomisation within the strata guards against bias. Another
method of sampling is systematic (for example interview at every 10th house). This
is usually much easier logistically than randomisation, and hence is often used.
However, it carries the danger of producing a biased sample, if there is variation in
the population that is correlated with your sampling. For example, if you interview
hunters every seventh day, you might always get them on market day, or always get
them on their day off, and hence your results will not be representative of their
weekly activities.
In qualitative research, the focus is not on randomisation to ensure representa-
tiveness. Instead, representativeness is ensured based on an understanding of the
system, which the researcher builds up during the study. The researcher may
actively target extreme cases, or cases that illustrate borderline situations.
