6
A . MACFADYEN
practical decisions such as distance from the laboratory and transport
facilities are involved here but a useful example of a comparison of this
kind is given by Finney (1946) which is also quoted by Healy (in press).
In the example of a wireworm survey conducted by Finney two different sample sizes were tried and in each case the mean and standard
error were calculated. By plotting slm against m and joining the points
by eye an expected value of s for any given value of m can be obtained.
Since the number of units required to give the same precision are in the
ratios of the squares of the slm ratios the relative advantages of using
different sample sizes can be assessed. I n this particular instance there
proved to be little advantage in using 2in diameter cores rather than
4in diameter ones.
Decisions about sample size will, of course be very much influenced by
whether or not abolute numbers are to be determined, that is, whether
the survey is of the “trophic” or “community” type as discussed in
Section 11. Most ecological work done under expedition conditions is of
the second kind and is often aimed a t detecting and delimiting characteristic species groupings and relating these to environmental factors.
Although it has quite commonly been the practice to count all specimens contained in each sample-unit, various expedients can be employed to derive quantitative information from samples without necessitating complete counts. I n continental Europe it is usual to estimate
numbers to the nearest order of magnitude or to employ an arbitrary
abundance scale as has been done by Gisin (1943) and Strenzke (1952)
although rather little use is made of these ratings. Presence or absence
can be recorded and used to calculate “frequency” (i.e., proportion of
sample-units in which a given species occurs). Frequency and mere presence are the basis of correlation tables (or “Trellis diagrams”) which are
used for community analysis in Scandinavia particularly, for example
in the work of Kontkanen (1950,1957). The recording of the simultaneous
occurrence of several species in this way provides a means of determining the extent to which their distributions are associated or complementary. The more modest studies of this kind are content with analysis
of correlation between two species a t a time, for example, the work of
Cole (1949). A recent development however has been the simultaneoua
correlation analysis of presence or absence of a large number of species
by Williams and Lambert (1959, 1960) with the aid of a computer. This
involves the complete sampling of a large gridded area and recording
species lists for each grid square. The squares are then classified on a
hierarchical system which separates those species groups whose distribution is least significantly associated using all possible combinations of
species. I n this way previously unsuspected correlations between vegetation pattern and environmental factors are detected.
A . MACFADYEN
practical decisions such as distance from the laboratory and transport
facilities are involved here but a useful example of a comparison of this
kind is given by Finney (1946) which is also quoted by Healy (in press).
In the example of a wireworm survey conducted by Finney two different sample sizes were tried and in each case the mean and standard
error were calculated. By plotting slm against m and joining the points
by eye an expected value of s for any given value of m can be obtained.
Since the number of units required to give the same precision are in the
ratios of the squares of the slm ratios the relative advantages of using
different sample sizes can be assessed. I n this particular instance there
proved to be little advantage in using 2in diameter cores rather than
4in diameter ones.
Decisions about sample size will, of course be very much influenced by
whether or not abolute numbers are to be determined, that is, whether
the survey is of the “trophic” or “community” type as discussed in
Section 11. Most ecological work done under expedition conditions is of
the second kind and is often aimed a t detecting and delimiting characteristic species groupings and relating these to environmental factors.
Although it has quite commonly been the practice to count all specimens contained in each sample-unit, various expedients can be employed to derive quantitative information from samples without necessitating complete counts. I n continental Europe it is usual to estimate
numbers to the nearest order of magnitude or to employ an arbitrary
abundance scale as has been done by Gisin (1943) and Strenzke (1952)
although rather little use is made of these ratings. Presence or absence
can be recorded and used to calculate “frequency” (i.e., proportion of
sample-units in which a given species occurs). Frequency and mere presence are the basis of correlation tables (or “Trellis diagrams”) which are
used for community analysis in Scandinavia particularly, for example
in the work of Kontkanen (1950,1957). The recording of the simultaneous
occurrence of several species in this way provides a means of determining the extent to which their distributions are associated or complementary. The more modest studies of this kind are content with analysis
of correlation between two species a t a time, for example, the work of
Cole (1949). A recent development however has been the simultaneoua
correlation analysis of presence or absence of a large number of species
by Williams and Lambert (1959, 1960) with the aid of a computer. This
involves the complete sampling of a large gridded area and recording
species lists for each grid square. The squares are then classified on a
hierarchical system which separates those species groups whose distribution is least significantly associated using all possible combinations of
species. I n this way previously unsuspected correlations between vegetation pattern and environmental factors are detected.