2. Inferring Threat from Scientific Collections
13
Figure 2.3. Power curves for three equations for a scenario in which the underlying
population declined linearly at 0.5, 1, 2, and 5% per time step. (a) Solow’s equation; (b)
Grimson’s equation.
improve in circumstances in which cells with more than a single observation are
relatively common.
The general shape of the stepped-decline function curves was similar to those
generated by the continuous-decline scenario (Fig. 2.4). The equations performed
better when the mean number of observations increased and when the change in
the status of the underlying population was more marked. A 100% reduction in
population size halfway through a 100-time-step sequence was much easier to
detect than was a 20% reduction.
Neither of the equations was effective at detecting stepped declines of less than
about 50%. There was little corresponding increase in power with an increase in
mean number of observations per time step at these levels of decline. When the
underlying population declined by more than 50% at the halfway point, the
equations were likely to detect a change if the mean number of observations per
time step was greater than about 0.25. The likelihood that a true change would be
detected approached 100% when the magnitude of the decline approached 100%
and the mean number of observations per time step was greater than about 0.25.
13
Figure 2.3. Power curves for three equations for a scenario in which the underlying
population declined linearly at 0.5, 1, 2, and 5% per time step. (a) Solow’s equation; (b)
Grimson’s equation.
improve in circumstances in which cells with more than a single observation are
relatively common.
The general shape of the stepped-decline function curves was similar to those
generated by the continuous-decline scenario (Fig. 2.4). The equations performed
better when the mean number of observations increased and when the change in
the status of the underlying population was more marked. A 100% reduction in
population size halfway through a 100-time-step sequence was much easier to
detect than was a 20% reduction.
Neither of the equations was effective at detecting stepped declines of less than
about 50%. There was little corresponding increase in power with an increase in
mean number of observations per time step at these levels of decline. When the
underlying population declined by more than 50% at the halfway point, the
equations were likely to detect a change if the mean number of observations per
time step was greater than about 0.25. The likelihood that a true change would be
detected approached 100% when the magnitude of the decline approached 100%
and the mean number of observations per time step was greater than about 0.25.