4.4 Statistics in ELA W A T
n
4.4
Statistics in ELAWAT
The contribution of statistics to ecosystem research consists primarily in making
ecological terms measurable, because what cannot be measured cannot be treated
statistically. Thereby, Applied Statistics represents a bridge between system properties on a conceptual level and the measurements on the level of investigation.
Stability properties of ecological systems in the sense of Grimm (1996) refer to
ecological variables (which are not necessarily measurable in the sense of the
present chapter) and their change related to disturbances. The objective is a comparison of states or developments after a disturbance to a reference state or reference dynamic, respectively. From the attempt to make terms measurable, the following questions arose in ELA W A T: Which components of the system, which
ecological variables, can be quantified? Which measurable variables characterize
their state? In which case are states to be regarded similar? Which distance between states lies within a reference dynamic? In which way can dissimilarity of
states be proven statistically') After the statistical analysis and the interpretation of
the results another bridge must be constructed. Statistical techniques help understanding ecological phenomena, they do not replace the elaboration of ecological
results. Statistical analysis can, at best, support but not replace ecologists' expertise
and creative reasoning.
For application in ecology a great number of, partly numerically demanding,
statistical techniques are offered. During ELA W AT mainly exploratory statistical
techniques were applied (e.g. Box-and-Whisker-Plots, MDS, variograms). Especially MDS, as a very flexible and evident technique, was generally accepted and
therefore frequently mentioned and applied in the case studies of the present
chapter. Also the other multivariate techniques presented in this chapter are of
great potential for the understanding of relations in ecosystems. But they were less
accepted. When test procedures were applied, these were mostly non-parametric
tests (e.g. U-Test, multivariate randomization tests).
Statistical techniques are sometimes selected with regard to the statistical software package available, but not with regard to statistical reasoning. This mechanism will often be inevitable for the scientist. Nevertheless, one should be aware of
it and consider this critically for the assessment of results. For ecologists this requires a good overview of statistical methods beyond those implemented in the
statistical software packages at hand. For statisticians it requires a fast transition
from research results into availabl~ software, as was done in ELA W A T.
A successful application of statistical models and techniques demands from
ecologists a basic knowledge in statistics and openness for new techniques. From
statisticians it demands the ability of translating statistical subjects into the language of ecology. It also requires the chance for personal exchange so that, as was
done in ELA W AT, ecological questions and statistical techniques can be explained
and discussed in common.
Statistics is only one of many ways of advancing scientific progress. Its potential ends when no measurements were made or data do not have the required quality. All the statistical techniques explicitly or implicitly assume a model. The part
n
4.4
Statistics in ELAWAT
The contribution of statistics to ecosystem research consists primarily in making
ecological terms measurable, because what cannot be measured cannot be treated
statistically. Thereby, Applied Statistics represents a bridge between system properties on a conceptual level and the measurements on the level of investigation.
Stability properties of ecological systems in the sense of Grimm (1996) refer to
ecological variables (which are not necessarily measurable in the sense of the
present chapter) and their change related to disturbances. The objective is a comparison of states or developments after a disturbance to a reference state or reference dynamic, respectively. From the attempt to make terms measurable, the following questions arose in ELA W A T: Which components of the system, which
ecological variables, can be quantified? Which measurable variables characterize
their state? In which case are states to be regarded similar? Which distance between states lies within a reference dynamic? In which way can dissimilarity of
states be proven statistically') After the statistical analysis and the interpretation of
the results another bridge must be constructed. Statistical techniques help understanding ecological phenomena, they do not replace the elaboration of ecological
results. Statistical analysis can, at best, support but not replace ecologists' expertise
and creative reasoning.
For application in ecology a great number of, partly numerically demanding,
statistical techniques are offered. During ELA W AT mainly exploratory statistical
techniques were applied (e.g. Box-and-Whisker-Plots, MDS, variograms). Especially MDS, as a very flexible and evident technique, was generally accepted and
therefore frequently mentioned and applied in the case studies of the present
chapter. Also the other multivariate techniques presented in this chapter are of
great potential for the understanding of relations in ecosystems. But they were less
accepted. When test procedures were applied, these were mostly non-parametric
tests (e.g. U-Test, multivariate randomization tests).
Statistical techniques are sometimes selected with regard to the statistical software package available, but not with regard to statistical reasoning. This mechanism will often be inevitable for the scientist. Nevertheless, one should be aware of
it and consider this critically for the assessment of results. For ecologists this requires a good overview of statistical methods beyond those implemented in the
statistical software packages at hand. For statisticians it requires a fast transition
from research results into availabl~ software, as was done in ELA W A T.
A successful application of statistical models and techniques demands from
ecologists a basic knowledge in statistics and openness for new techniques. From
statisticians it demands the ability of translating statistical subjects into the language of ecology. It also requires the chance for personal exchange so that, as was
done in ELA W AT, ecological questions and statistical techniques can be explained
and discussed in common.
Statistics is only one of many ways of advancing scientific progress. Its potential ends when no measurements were made or data do not have the required quality. All the statistical techniques explicitly or implicitly assume a model. The part