Environmental Engineering Reference
In-Depth Information
geographic range of the study to remove the nuisance effect of varying environmen-
tal conditions to ensure that the effect of burning was determined.
Although it is important to control for as many variables as possible in a study
design, there is a limit to the number of controllable or measurable variables for
most ecological studies. This is true for both field and laboratory studies. Fre-
quently, it is not possible to identify or recognize all of the potential variables
affecting a dependent variable. At times, it is difficult, economically unfeasible, or
impossible to measure certain variables. Therefore, it is often appropriate to
identify and measure proxy (i.e., correlated) variables that can serve as an index
to the variable of interest. Finally, there is a statistical limit to the number of
variables that can be addressed through study design as well, primarily because
sample size relative to the number of variables dictates the potential analyses. Such
uncontrollable variables are typically assumed to be random with the same effect
across all samples and controlled variables and thus accounted for in the appropriate
experimental design.
Proxy variables can take many forms and do not have to be directly related to the
variable of interest. For example, Mackay et al. (
2007
) measured soil moisture,
which affects plant root water availability, to use as a proxy variable for detecting
water stress. Pfeiffer (
2007
) used wetland and other surface water area as a proxy
variable for the location of spatially clustered wild bird infection in a study of the
influence of wild birds and risk from H5N1 highly-pathogenic avian influenza. In a
review study, Elser et al. (
2007
) identified several variables that were correlated
(i.e., proxy) with standing biomass of autotrophs including chlorophyll concentra-
tion, ash-free dry mass, carbon mass, biovolume, percent cover, and primary
production. They then used a meta-analysis combining results from nitrogen and
phosphorus enrichment studies measuring standing biomass and proxies to evaluate
nutrient limitation in freshwater, marine, and terrestrial ecosystems. Kantrud and
Newton (
1996
) used the amount of cropland in a wetland watershed as a proxy for
their quality. Use of proxy variables is common in wetland studies due to the
difficultly in measurements of many ecological characteristics; however, one
must be somewhat reserved in stating conclusions using proxy variables unless
certainty exists relative to the strength of the relationship between a proxy variable
and the variable of interest.
1.6 Conceptual Model and Variable Selection
A common approach to determine which variables to measure for testing competing
hypotheses is to transform a conceptual hypothesis to a conceptual model that
describes the response of a dependent variable to a set of independent variables.
The conceptual model forms the basis of an appropriate statistical model with
defined components. The statistical model can be formulated as Y (dependent
variable) being a function (
F
) of some fixed and random independent variables or
factors. For example, if one is measuring the total nitrogen load in a wetland, Y
