2. Power analysis on real ibex trends in GPNP
If the complete trend observed in the ibex count data in GPNP was
significant (86.9% of the time intervals), sample counts were able to
correctly detect its direction with a sufficient statistical power (fig.
5a) using a number of sectors between 5 (for a 20-year trend) and 28
(for a 5-year trend). To correctly measure the magnitude of the trend, a
higher number of sectors was required (fig.5b). The complete trend was
detected with an error lower than 2% and in at least 80% of the cases
if 10 (for a 20-years trend) to 33 (for a 5-years trend) sectors were
surveyed. To detect a trend over 10 years, 15 sectors were needed to
correctly assess the direction of the trend and 22 sectors to also
detect its magnitude, while for a 20-year trend 10 sectors were
sufficient to accurately quantify the trend. Sampling more sectors
reduced the number of false, missed and opposing trends (fig.6). Indeed,
counting in the full area compared to half of sectors led to false
trends declining from 14% to 6%, missed trends from 10% to 3% and
opposing trends from 5% to 0%.