Fróðskaparrit - 31.12.2000, Page 105
TITTLEIKIN AV VANLIGASTU FØROYSKU DýRA- OG TARASLØGUNUM
A KLETTASTROND I MUN TIL YMISK UMHVØRVISVIðURSKIFTI
109
at a 5% level. The marginal effect of each
variable, which includes effects due to cor-
relation with other variables, was deter-
mined using the variance extracted by CCA
with the given variable entered as the only
environmental variable. The unique effect
of each variable, which only includes vari-
ance that cannot be extracted by the other
variables, was determined using the vari-
ance extracted by partial CCA, in which the
given variable was entered as the only envi-
ronmental variable and all other variables
were entered as co-variables.
In order to explore the variation unac-
counted for by the CCA, Partial Correspon-
dence Analysis (partial CA) was used with
the previously selected environmental vari-
ables entered as co-variables. In order to
view all variation in the species data inde-
pendently of the estimated environmental
Table 3. Canonical Correspondence Analysis (CCA)for rocky shore communities in the Faroe Islands,
showing eigenvalues, extracted percentage variance ofspecies data and species-environment relations for the
firstfour canonical axes. Inertia: weighted variance of species data. Marginal effect: inertia explained by each
variable if selected as the only variable. Unique effect: inertia explained by each variable if the other variables
are entered as co-variables. The variables of importance were selected by the ‘forward selection ’ procedure, by
which the vdriables were selected one by one in sequence and included if significant at a 5% level. The data
comprise 22 species at 159 sites with tidal amplitude larger than 0.4 m.
Axis 1 2 3 4
Eigenvalue 0.262 0.018 0.009 0.007
Cumulative extracted variance (%) of species data 29.2 31.3 32.3 33.1
- of species - environment relation 87.6 93.8 96.9 99.2
Inertia %
Total inertia (sum of all unconstrained eigenvalues) 0.898 100
Variance explained by environmental variables (sum of all canonical eigenvalues) 0.299 33
Unexplained variance 0.599 67
Marginal effect Unique effect Forward selection
(percentage of (percentage of (percentage of
Variable total inertia) total inertia) total inertia added)
Exposure 19 * 8 * 19 *
Substrate 14* 7 * 9*
Fjord index 13 * 2 * 3 *
Current 2 * i * 1 *
Slope 2 * i -
Tide 1 * 2 * 2*
Local protection 1 0 -
Aspect 1 1 -
* Significant at a 5% level (Monte Carlo permutation test, 999 unrestricted permutations)