4.3 Statistical Methods
65
Results: There is one situation in the first part and one situation in the second
part of the series in which the series does not increase or decrease, respectively, as
is assumed under the null hypothesis. This number of situations is the test statistic
for the Moore-Wallis test of first differences and it corresponds to a p-value of
0.042 (Bortz et a!. 1990, Table 40).
The U-Test of the median values from summer against winter results in a pvalue of 0.0275 (Bortz et a!. 1990, Table 6).
Both test results indicate that the quotient of Mn and AlP, differs over the year.
4.3.1.9
Multivariate Temporal Processes
Questions: What is the temporal development of sediment chemistry after a disturbance? What is the temporal development of occurrence and abundance of zooplankton during a year') In what way does recolonization by microphythobenthos,
nematodes and macrofauna develop after disturbance?
Multivariate measurements cannot, in general, be graphically represented as
points in two dimensions without loss of information. Multivariate series of measurements are difficult to represent graphically, because usually it will not be possible to find a ranking. A solution is to change over to dissimilarities between pairs
of measurements followed by MDS (compare Sect. 4.3.1.3).
If there is a valid configuration in one dimension, it will be possible to rank the
measurements and represent them as usual in a time diagram. If there is a valid
configuration in two dimensions, it will be possible to represent time linking the
points by arrows.
Another possibility is the choice of a reference point in time, to which all the
others are compared. In the time diagram, dissimilarity between the respecti ve
measurement and the reference measurement is shown.
4.3.1.10
Multivariate Temporal Processes Along a Transect
Questions: In what way do for example chemical, sedimentological or biological
variables develop along a transect from a mussel bed to a sandflat during one year?
Frequently, for the data analysis of such a question, temporal and spatial development are separated from each other and analysed one by one. This results in statements concerning the development along the transect, averaged over time, and
statements concerning the temporal development, averaged over the transect stations. Simultaneous statements about the spatial and temporal development are
achieved by a joint analysis of temporal and spatial multivariate measurements.
4.3.1.11
Case Study: Geochemical Variables Along a Transect During one
Year
Data provided by A. Hild.
Question: Which factor is more important for the structure of geochemical variables along a transect: seasonal difference or spatial distance?
65
Results: There is one situation in the first part and one situation in the second
part of the series in which the series does not increase or decrease, respectively, as
is assumed under the null hypothesis. This number of situations is the test statistic
for the Moore-Wallis test of first differences and it corresponds to a p-value of
0.042 (Bortz et a!. 1990, Table 40).
The U-Test of the median values from summer against winter results in a pvalue of 0.0275 (Bortz et a!. 1990, Table 6).
Both test results indicate that the quotient of Mn and AlP, differs over the year.
4.3.1.9
Multivariate Temporal Processes
Questions: What is the temporal development of sediment chemistry after a disturbance? What is the temporal development of occurrence and abundance of zooplankton during a year') In what way does recolonization by microphythobenthos,
nematodes and macrofauna develop after disturbance?
Multivariate measurements cannot, in general, be graphically represented as
points in two dimensions without loss of information. Multivariate series of measurements are difficult to represent graphically, because usually it will not be possible to find a ranking. A solution is to change over to dissimilarities between pairs
of measurements followed by MDS (compare Sect. 4.3.1.3).
If there is a valid configuration in one dimension, it will be possible to rank the
measurements and represent them as usual in a time diagram. If there is a valid
configuration in two dimensions, it will be possible to represent time linking the
points by arrows.
Another possibility is the choice of a reference point in time, to which all the
others are compared. In the time diagram, dissimilarity between the respecti ve
measurement and the reference measurement is shown.
4.3.1.10
Multivariate Temporal Processes Along a Transect
Questions: In what way do for example chemical, sedimentological or biological
variables develop along a transect from a mussel bed to a sandflat during one year?
Frequently, for the data analysis of such a question, temporal and spatial development are separated from each other and analysed one by one. This results in statements concerning the development along the transect, averaged over time, and
statements concerning the temporal development, averaged over the transect stations. Simultaneous statements about the spatial and temporal development are
achieved by a joint analysis of temporal and spatial multivariate measurements.
4.3.1.11
Case Study: Geochemical Variables Along a Transect During one
Year
Data provided by A. Hild.
Question: Which factor is more important for the structure of geochemical variables along a transect: seasonal difference or spatial distance?