30
Brendan Moyle
distributed errors. However, the probabilistic nonlinear relation between the explanatory and dependent variables means that least-squares measures are inappropriate estimators of goodness of fit.
In this study, the logistic and probit models (based on logit and probit link
functions) are used to relate extinction data to morphological and other attributes,
in an effort to predict the attributes of species that predispose them to extinction
risk.
Correlates of Extinction Risk: The Case of New Zealand
Terrestial Birds
The New Zealand avifauna has suffered important losses. Human settlement
triggered a wave of deforestation and introductions of exotic mammalian pests.
This was associated with the extinction of many bird species (Gill 1991). For
many species, extinction in the period before European settlement means that
there is little reliable information on population densities and distribution (East
and Williams 1984; Atkinson and Millener 1991).
These extinctions are amenable to analysis with logistic regression. The dependent variable is the current status of the species: either extinct or extant. Survey
data and the details of the ecology of most species are scant. The majority of the
information about these species is morphological.
Determining the vulnerability to extinction of these various species may be of
interest for a number of reasons. From a strictly ecological perspective, understanding why one species rather than another is extinct may be useful in explaining the pattern of observed extinctions. From a policy perspective, an understanding of what constitutes the risk factors facing New Zealand birds may aid
conservation planning. Such a model may also complement field studies or anticipate the decline of other species.
In this study, only terrestrial birds were included (Mills and Williams 1984).
Aquatic and marine birds are likely to be less vulnerable to some of the landscape
changes associated with human settlement. For instance, deforestation may be
presumed to have a greater impact on the terrestrial birds rather than the marine or
aquatic. Two terrestrial species (the Kea [Nestor notabilis] and the Rock Wren
[Xenicus gilviventris]) that had an alpine rather than lowland distribution were
also excluded from the set. Again, the rationale was to focus on a narrow landscape. Hence the study is really concerned with those species that were distributed
over forests and their terrestrial margins. This generated a sample of 80 taxa.
Model Specification
The dependent variable was the current status of each species: extinct or extant.
In this study, species that were recognized in 1984 as endangered (Mills and
Williams 1984) were classified as extinct because the persistence of these species
was attributable to human efforts at preservation rather than to any resilience of
Brendan Moyle
distributed errors. However, the probabilistic nonlinear relation between the explanatory and dependent variables means that least-squares measures are inappropriate estimators of goodness of fit.
In this study, the logistic and probit models (based on logit and probit link
functions) are used to relate extinction data to morphological and other attributes,
in an effort to predict the attributes of species that predispose them to extinction
risk.
Correlates of Extinction Risk: The Case of New Zealand
Terrestial Birds
The New Zealand avifauna has suffered important losses. Human settlement
triggered a wave of deforestation and introductions of exotic mammalian pests.
This was associated with the extinction of many bird species (Gill 1991). For
many species, extinction in the period before European settlement means that
there is little reliable information on population densities and distribution (East
and Williams 1984; Atkinson and Millener 1991).
These extinctions are amenable to analysis with logistic regression. The dependent variable is the current status of the species: either extinct or extant. Survey
data and the details of the ecology of most species are scant. The majority of the
information about these species is morphological.
Determining the vulnerability to extinction of these various species may be of
interest for a number of reasons. From a strictly ecological perspective, understanding why one species rather than another is extinct may be useful in explaining the pattern of observed extinctions. From a policy perspective, an understanding of what constitutes the risk factors facing New Zealand birds may aid
conservation planning. Such a model may also complement field studies or anticipate the decline of other species.
In this study, only terrestrial birds were included (Mills and Williams 1984).
Aquatic and marine birds are likely to be less vulnerable to some of the landscape
changes associated with human settlement. For instance, deforestation may be
presumed to have a greater impact on the terrestrial birds rather than the marine or
aquatic. Two terrestrial species (the Kea [Nestor notabilis] and the Rock Wren
[Xenicus gilviventris]) that had an alpine rather than lowland distribution were
also excluded from the set. Again, the rationale was to focus on a narrow landscape. Hence the study is really concerned with those species that were distributed
over forests and their terrestrial margins. This generated a sample of 80 taxa.
Model Specification
The dependent variable was the current status of each species: extinct or extant.
In this study, species that were recognized in 1984 as endangered (Mills and
Williams 1984) were classified as extinct because the persistence of these species
was attributable to human efforts at preservation rather than to any resilience of