[0.306; 0.758]. Therefore, there is a large probability that the true IndVal values for
these two species reside somewhere within the common part of these ranges, i.e.,
[0.391; 0.758], but there is no way of determining that one value is larger than the
other.
4.11.3 Correlation-Type Indices
As mentioned above, correlation indices were devised to help identify the ecological
preferences of species among a set of groups. De Cáceres and Legendre (2009)
pointed out that this approach is “probably more useful [for this purpose] than the
indicator value approach, because the former naturally allows the detection of
negative preferences”. Beware, however, of the problems relative to the absence
of species from a set of sites. The absence of a species may be due to different
reasons in each site, and thus it should not be interpreted in ecological terms without
due caution (see Sect. 3.2.2). Species with high ecological preferences for a given
group of sites representing well-defined ecological conditions are often called
“diagnostic” species in plant ecology and are useful to identify vegetation types in
field surveys (Chytrý et al. 2002 in De Cáceres and Legendre 2009).
The simplest correlation-type index for presence-absence data is called Pearson’s
ϕ (phi) coefficient of association (Chytrý et al. 2002 in De Cáceres and Legendre
2009). It consists in the correlation between two binary vectors. For all sites, one
vector gives the presence or absence of the species, and the other indicates that the
site belongs (1) or not (0) to a given group. With quantitative (abundance) data, the
phi coefficient is called the point biserial correlation coefficient and a vector of
abundances replaces the presence-absence vector. In function strassoc(), the
default indicator measure is the phi coefficient, requested by func ¼ "r", but an
"r.g" option is available that corrects the measure for unequal group sizes. We
recommend the latter option, following De Cáceres’ indicspecies tutorial.
Let us compute the phi coefficient (corrected for unequal group sizes) on the fish
abundance data.
# Phi correlation index
iva.phi <- multipatt(
spe,
grps,
func = "r.g",
max.order = 2,
control = how(nperm = 999)
)
summary(iva.phi)
The results of this analysis look quite similar to that of the IndVal analysis. Note
that the statistical tests highlight the highest positive phi values (see the summary).
Therefore, it is natural that both approaches (species indicator values and
4.11 Indicator Species
125
these two species reside somewhere within the common part of these ranges, i.e.,
[0.391; 0.758], but there is no way of determining that one value is larger than the
other.
4.11.3 Correlation-Type Indices
As mentioned above, correlation indices were devised to help identify the ecological
preferences of species among a set of groups. De Cáceres and Legendre (2009)
pointed out that this approach is “probably more useful [for this purpose] than the
indicator value approach, because the former naturally allows the detection of
negative preferences”. Beware, however, of the problems relative to the absence
of species from a set of sites. The absence of a species may be due to different
reasons in each site, and thus it should not be interpreted in ecological terms without
due caution (see Sect. 3.2.2). Species with high ecological preferences for a given
group of sites representing well-defined ecological conditions are often called
“diagnostic” species in plant ecology and are useful to identify vegetation types in
field surveys (Chytrý et al. 2002 in De Cáceres and Legendre 2009).
The simplest correlation-type index for presence-absence data is called Pearson’s
ϕ (phi) coefficient of association (Chytrý et al. 2002 in De Cáceres and Legendre
2009). It consists in the correlation between two binary vectors. For all sites, one
vector gives the presence or absence of the species, and the other indicates that the
site belongs (1) or not (0) to a given group. With quantitative (abundance) data, the
phi coefficient is called the point biserial correlation coefficient and a vector of
abundances replaces the presence-absence vector. In function strassoc(), the
default indicator measure is the phi coefficient, requested by func ¼ "r", but an
"r.g" option is available that corrects the measure for unequal group sizes. We
recommend the latter option, following De Cáceres’ indicspecies tutorial.
Let us compute the phi coefficient (corrected for unequal group sizes) on the fish
abundance data.
# Phi correlation index
iva.phi <- multipatt(
spe,
grps,
func = "r.g",
max.order = 2,
control = how(nperm = 999)
)
summary(iva.phi)
The results of this analysis look quite similar to that of the IndVal analysis. Note
that the statistical tests highlight the highest positive phi values (see the summary).
Therefore, it is natural that both approaches (species indicator values and
4.11 Indicator Species
125