60
J . M. LAMBERT AND M. B . DALE
(1964), whose latest contributions between them provide a comprehensive bibliography of the relevant literature to date. With such
modern surveys available, we see no point in repetition of similar
material here. Instead, we shall attempt to examine the present situation more particularly in terms of current concepts which may be
involved in efforts to put the study of vegetation oq an increasingly
objective basis; and this in turn will require the close examination of
certain basic assumptions, with regard both to the nature of the data
to be analysed and to the statistical methods which can be employed.
B. D E F I N I T I O N OF TERMS
1. The DeJinition of “Statistics”
The modern usage of the term “statistics” has tended to restrict its
meaning to probabilistic studies, i.e. to studies of problems involving,
a degree of uncertainty and, more specifically, to those involving estimation of parameters and testing of hypotheses previously erected. However, the word has an older - though now dubiously respectable -
meaning, derived from its original use for the description of “state”
data (i.e. political and economic data) and thereby including nonprobabilistic methods for data simplification and generation of hypotheses. Although both types of method have a part to play in ecological
work, the non-probabilistic techniques are in fact more generally
relevant to the kind of empirical situation usually met with in the field;
and although this aspect has been largely rejected from modern
statistical terminology, there is as yet no convenient alternative word
available t o cover it. In the present contribution, therefore, we shall
deliberately revert to the older and wider definition of statistics covering
both aspects, rather than use still more ambiguous terms like “numerical”
and “quantitative” methods.
This wider definition thus allows us to include two important primary
functions beyond those appertaining to more orthodox techniques. The
first concerns the reduction, subject to some efficiency criterion, of a
large and complex mass of data into a more accessible and convenient
form; such simplification aims only at allowing the investigator to
reduce the amount of information to be handled for future examination.
The second - and generally more useful - function is that of hypothesis-generation. Hypotheses normally postulate underlying causal
factors which may be thought of as having generated the original data
in all their complexity, and there is no a priori reason why they should
not be developed from the original data. In complex situations, there
may be so many variables that the whole pattern cannot be intuitively
grasped; if, however, the data can be so simplified that their internal
J . M. LAMBERT AND M. B . DALE
(1964), whose latest contributions between them provide a comprehensive bibliography of the relevant literature to date. With such
modern surveys available, we see no point in repetition of similar
material here. Instead, we shall attempt to examine the present situation more particularly in terms of current concepts which may be
involved in efforts to put the study of vegetation oq an increasingly
objective basis; and this in turn will require the close examination of
certain basic assumptions, with regard both to the nature of the data
to be analysed and to the statistical methods which can be employed.
B. D E F I N I T I O N OF TERMS
1. The DeJinition of “Statistics”
The modern usage of the term “statistics” has tended to restrict its
meaning to probabilistic studies, i.e. to studies of problems involving,
a degree of uncertainty and, more specifically, to those involving estimation of parameters and testing of hypotheses previously erected. However, the word has an older - though now dubiously respectable -
meaning, derived from its original use for the description of “state”
data (i.e. political and economic data) and thereby including nonprobabilistic methods for data simplification and generation of hypotheses. Although both types of method have a part to play in ecological
work, the non-probabilistic techniques are in fact more generally
relevant to the kind of empirical situation usually met with in the field;
and although this aspect has been largely rejected from modern
statistical terminology, there is as yet no convenient alternative word
available t o cover it. In the present contribution, therefore, we shall
deliberately revert to the older and wider definition of statistics covering
both aspects, rather than use still more ambiguous terms like “numerical”
and “quantitative” methods.
This wider definition thus allows us to include two important primary
functions beyond those appertaining to more orthodox techniques. The
first concerns the reduction, subject to some efficiency criterion, of a
large and complex mass of data into a more accessible and convenient
form; such simplification aims only at allowing the investigator to
reduce the amount of information to be handled for future examination.
The second - and generally more useful - function is that of hypothesis-generation. Hypotheses normally postulate underlying causal
factors which may be thought of as having generated the original data
in all their complexity, and there is no a priori reason why they should
not be developed from the original data. In complex situations, there
may be so many variables that the whole pattern cannot be intuitively
grasped; if, however, the data can be so simplified that their internal