THE USE O F STATISTICS IN PHYTOSOCIOLOCIY
91
the available information in as much detail as possible, “abstraction” in
contrast seeks only to retain that which is important for the purpose
in hand and to eliminate the rest. The function of statistics is not to
describe, but to assist the investigator in the process of abstraction;
moreover, by indicating the more significant features in the data, it
may also assist in the process of hypothesization.
If the function of phytosociology is only “to describe vegetation and
to discover and define problems”, then it is scarcely worthy to be
recognized as a science at all. We prefer to think of phytosociology in a
wider context, i.e. as scientific investigation rather than a description
of vegetational relationships. The essence of scientific work lies in the
erection and testing of hypotheses, from which predictions can be made.
The function of primary survey is to establish an initial pattern;
this is then used as the basis for the erection of hypotheses, the testing
of which requires the collection of more information. The hypothesis
to be tested must not inadvertently be incorporated into the pattern,
and if the object of the exercise is to determine plant/habitat relationships, then the one or the other must be excluded from the pattern
used for hypothesization. However, once correlations have been
established and independently tested, then abstractions can be made
from the joint information for use in other contexts.
Where phytosociological (i.e. plantlsite) relationships are used as a
means of entry into ecological problems, rather than just as a means of
discovering and defining them, then we have already suggested (p. 65)
that the most practical approach is to establish the phytosociological
pattern first. Given that no assumption of discrete communities is
made (p. 89) and that the sampling system is predetermined on a
systematic basis, the question now arises as to whether it is more
efficient to determine the overall pattern by “wholistic” methods (i.e.
simultaneous assessment of the total information) or by “sequential”
means.
If we understand Poore aright (1962, p. 38), his “method of successive
approximation” involves the sequential assessment of information.
However, it is by no means clear from his description of the method
whether it refers only to the process of progressively adjusting hypotheses in the light of sequential observation, or whether in fact he is
also concerned with methods of testing hypotheses or deriving estimates
by the use of samples in sequence so that a decision can be taken as
soon as the samples provide sufficient information according to some
criterion : the latter approach has some parallel in “sequential” statistical methods and might be justifiable in certain circumstances. In his
advocacy of the method, Poore writes (Zoc. cit., p. 39): “The method of
successive approximation would appear to be the most economical way
91
the available information in as much detail as possible, “abstraction” in
contrast seeks only to retain that which is important for the purpose
in hand and to eliminate the rest. The function of statistics is not to
describe, but to assist the investigator in the process of abstraction;
moreover, by indicating the more significant features in the data, it
may also assist in the process of hypothesization.
If the function of phytosociology is only “to describe vegetation and
to discover and define problems”, then it is scarcely worthy to be
recognized as a science at all. We prefer to think of phytosociology in a
wider context, i.e. as scientific investigation rather than a description
of vegetational relationships. The essence of scientific work lies in the
erection and testing of hypotheses, from which predictions can be made.
The function of primary survey is to establish an initial pattern;
this is then used as the basis for the erection of hypotheses, the testing
of which requires the collection of more information. The hypothesis
to be tested must not inadvertently be incorporated into the pattern,
and if the object of the exercise is to determine plant/habitat relationships, then the one or the other must be excluded from the pattern
used for hypothesization. However, once correlations have been
established and independently tested, then abstractions can be made
from the joint information for use in other contexts.
Where phytosociological (i.e. plantlsite) relationships are used as a
means of entry into ecological problems, rather than just as a means of
discovering and defining them, then we have already suggested (p. 65)
that the most practical approach is to establish the phytosociological
pattern first. Given that no assumption of discrete communities is
made (p. 89) and that the sampling system is predetermined on a
systematic basis, the question now arises as to whether it is more
efficient to determine the overall pattern by “wholistic” methods (i.e.
simultaneous assessment of the total information) or by “sequential”
means.
If we understand Poore aright (1962, p. 38), his “method of successive
approximation” involves the sequential assessment of information.
However, it is by no means clear from his description of the method
whether it refers only to the process of progressively adjusting hypotheses in the light of sequential observation, or whether in fact he is
also concerned with methods of testing hypotheses or deriving estimates
by the use of samples in sequence so that a decision can be taken as
soon as the samples provide sufficient information according to some
criterion : the latter approach has some parallel in “sequential” statistical methods and might be justifiable in certain circumstances. In his
advocacy of the method, Poore writes (Zoc. cit., p. 39): “The method of
successive approximation would appear to be the most economical way