4.2 TermInology
53
mental frame for data analysis. They are necessary for assessing the statistical
properties of the results. Statistical techniques serve as tools for planning and analysing experiments. In the circular course of research from questions over answers
to new questions, applied statistics is directly linked to the investigation (observation or experiment). The aim of applying statistics in ecology is understanding
phenomena by interpreting data.
An attempt to bring mathematical modelling and statistical data analysis closer
together was made by Richter & Sbndgerath (1990). Statistical techniques can be
applied to analyse simulated data from ecological models (Samietz & Berger
1997). In general, however, mathematical modelling and statistics are treated separately in theory and application.
Statistical techniques can be classified under different aspects. If the question is
the result of a quantitative pilot study and much is known already about the variable under consideration (e.g. its statistical distribution), then the question can be
formulated as a statistical null hypothesis with alternative hypothesis. This is part
of statistical inference. One important principle in statistical inference is the formulation of the null hypothesis before sampling. On the contrary, in descriptive
statistics, the description of data is most important. Exploratory statistics provides
tools for finding unknown relations, structures and particularities in the data with
the aim of generating models and hypotheses (Bock 1980).
Ecosystem research is characterised by the search for and the analysis of relations between different parts of a system. Regarding data analysis in this context
means that it is rather the exception than the rule that a single variable is to be
analysed (univariate statistics). Thus, multivariate techniques are of special importance.
Also, the spatial and temporal scales play an important role in ecosystem research. Observations or measurements are often not independent, as required by
many statistical techniques. Rather, it is their temporal or spatial relation which is
to be investigated.
These different approaches show that there cannot be an overall rule for the application of statistical techniques. The choice of an adequate technique depends
rather on the question at hand and of the kind of data sampled. Underwood (1994)
gives examples showing in which respect ecological data do not fulfil the requirements of traditional statistical techniques.
4.2
Terminology
A variety of statistical techniques are currently used in ecology. Depending on
question and research focus, different models and techniques are preferred. There
is an extensive, sometimes confusing vocabulary, which has not been standardized.
Thus, depending on the context, there may be different names for the same statistical technique. The other way round, the same name may be used for different
techniques, depending on the context. In different disciplines, terms may have
different meanings, which may complicate the understanding. In this section an
attempt is made to set out statistical paradigms and terms in order to prevent misunderstandings as much as possible.
53
mental frame for data analysis. They are necessary for assessing the statistical
properties of the results. Statistical techniques serve as tools for planning and analysing experiments. In the circular course of research from questions over answers
to new questions, applied statistics is directly linked to the investigation (observation or experiment). The aim of applying statistics in ecology is understanding
phenomena by interpreting data.
An attempt to bring mathematical modelling and statistical data analysis closer
together was made by Richter & Sbndgerath (1990). Statistical techniques can be
applied to analyse simulated data from ecological models (Samietz & Berger
1997). In general, however, mathematical modelling and statistics are treated separately in theory and application.
Statistical techniques can be classified under different aspects. If the question is
the result of a quantitative pilot study and much is known already about the variable under consideration (e.g. its statistical distribution), then the question can be
formulated as a statistical null hypothesis with alternative hypothesis. This is part
of statistical inference. One important principle in statistical inference is the formulation of the null hypothesis before sampling. On the contrary, in descriptive
statistics, the description of data is most important. Exploratory statistics provides
tools for finding unknown relations, structures and particularities in the data with
the aim of generating models and hypotheses (Bock 1980).
Ecosystem research is characterised by the search for and the analysis of relations between different parts of a system. Regarding data analysis in this context
means that it is rather the exception than the rule that a single variable is to be
analysed (univariate statistics). Thus, multivariate techniques are of special importance.
Also, the spatial and temporal scales play an important role in ecosystem research. Observations or measurements are often not independent, as required by
many statistical techniques. Rather, it is their temporal or spatial relation which is
to be investigated.
These different approaches show that there cannot be an overall rule for the application of statistical techniques. The choice of an adequate technique depends
rather on the question at hand and of the kind of data sampled. Underwood (1994)
gives examples showing in which respect ecological data do not fulfil the requirements of traditional statistical techniques.
4.2
Terminology
A variety of statistical techniques are currently used in ecology. Depending on
question and research focus, different models and techniques are preferred. There
is an extensive, sometimes confusing vocabulary, which has not been standardized.
Thus, depending on the context, there may be different names for the same statistical technique. The other way round, the same name may be used for different
techniques, depending on the context. In different disciplines, terms may have
different meanings, which may complicate the understanding. In this section an
attempt is made to set out statistical paradigms and terms in order to prevent misunderstandings as much as possible.