Study Design and Logistics - Wetland Techniques - page 8

of terrestrial buffer strips of 100 m in width filters 90 % of suspended material
present in precipitation run-off; therefore, creating buffer strips around wetlands
will minimize or eliminate the potential of a wetland filling due to accumulated
terrestrial sediments. Deduction is reasoning from the general (premise) to a
specific event and includes the development of testable explicit predictions under
a specific hypothesis to explain observations from a natural system. An example of
deductive reasoning would be if avian body mass varied in relation to environmen-
tal conditions, then an investigator may predict that average body mass would
decline as temperature declines or number of days below freezing increases. The
use of retroduction involves the relatively subjective attribution of an underlying
cause to an observed pattern and is a common occurrence in the discussion section
of scientific papers. Conclusions based on retroduction should be considered
hypotheses that require additional testing. Most conclusions from scientific studies
are the result of retroduction and induction, which Romesburg ( 1981 ) lamented
does not result in gaining reliable knowledge as one would when using deduction.
This hypothetico-deductive approach was advocated by Romesburg ( 1981 ) as the
preferred approach for study of natural systems. Guthery ( 2008 ) described the
approach as the classical method to test a hypothesis by deducing events or
relationships that should be observed under experimentation if the hypothesis was
true. Specific, sound hypotheses result in efficient, organized, and goal-oriented
studies that minimize uncertainty in results.
Most commonly, hypotheses are stated in terms of treatment effects. Historically,
treatment effects were based on potential outcomes of statistical tests. Classical
statistical, “null” hypotheses are usually depicted in shorthand as H o : and generally
specify that “no difference” exists among treatments. Whereas the “alternative”
hypothesis H 1 : can be more specific and takes several forms related to the stated
hypotheses but usually is a statement that a difference exists among treatments.
Alternative hypotheses may also define a magnitude and direction of a difference. It
is important to remember that studies can reject or disprove a null hypothesis, but a
hypothesis cannot be considered proven when a null hypothesis is not rejected.
For example, a wetland manager can hypothesize that cattails ( Typha spp.) will be
completely eradicated from a wetland due to herbicide treatment, but the presence of
a single cattail would cause the hypothesis to be rejected. For the hypothesis to be
“accepted,” every plant would have to be identified to prove the null hypothesis and
it is impossible to be completely certain that all plants have been correctly identified.
Therefore, basic to scientific endeavor is the use of study design to formalize the
effort to disprove hypotheses rather than prove them (Peirce 1958 ).
Recently, use of the null statistical hypotheses has been dismissed as uninforma-
tive and nonproductive in scientific endeavors because use of a statistical test that
only declares whether a difference exists or not exists between treatment means is
relatively uninformative (e.g., Johnson 1999 ; Cherry 1999 ; Anderson 2008 ;
Guthery 2008 ). Instead, multiple alternative hypotheses should be developed that
are specific and depict cause and effect, measurable predictions, or explicit
outcomes that can be tested using data (Platt 1964 ; Romesburg 1981 ; Morrison
et al. 2001 ). In a study design, this can be relative to study objectives as long as one