Geoscience Reference
In-Depth Information
in the sample. In practice, the inclusion probabilities of sites not selected in
the sample are not needed, but they need to be knowable if the sample is
to meet the definition of a probability sample. To produce knowable inclu-
sion probabilities, the algorithm selecting the sample must be repeatable and
involve some sort of stochastic component.
A working definition of scientific design is then any sampling algorithm
that results in a probability sample. In this view, a probability sample and
scientific design are synonymous. The term
scientiic design
is used in what
follows because it is common and is often more natural in communications.
In general, nonscientific designs are simply those that do not include a prob-
ability sampling component. Under this definition, the universe of nonscien-
tific designs is vast and varied. Generally, nonscientific designs either are not
repeatable or do not involve some sort of random component.
The perils of nonscientific designs are well known (McDonald, 2003; Olsen
et al
.
, 1999; Edwards, 1998). Yet, nonscientific designs remain common and
overtly attractive. Nonscientific designs are attractive primarily because they
are easier, quicker, and less expensive to implement than scientific designs.
To illustrate the point that nonscientific designs are common, three types are
described next.
10.2.2 Judgment Samples
Judgment samples are those for which researchers, usually familiar with the
resource, make a judgment or decision regarding the best places to locate sam-
ple sites. To justify this, researchers may represent that they placed sample
sites in places that are representative, where the most change is anticipated,
in critical habitat, where the impact will be, or where interesting things will
happen. In doing so, the researchers are making judgments or assumptions
about relationships among the study's parameters that may be true or false.
In limited cases, when a research study's objectives are extremely narrow or if
the study is a true pilot study, judgment samples may be acceptable. But, even
in these cases, the assumptions and judgments made during site selection
need to be acknowledged and tested. The real danger of judgment samples is
that the researcher's assumptions are false, or only partly true, and the study
results are biased.
10.2.3 Haphazard Samples
Haphazard samples are those collected without a defined or quantifiable
sampling plan. Under haphazard sampling, the placement of sample sites
is not usually planned in advance. Many times, haphazard samples are
placed and taken where and when field crews have extra time after perform-
ing other duties. In many cases, field crews are responsible for determining
sample site placement while they are in the field, and unless crews are aware
of the pitfalls, sample sites are often placed wherever the crew happens to
