Multiple counts can be obtained by counting the same study site repeatedly
in the same season, or by counting multiple study sites once. The first option
tells us about temporal variation at sites within a season, the second about
spatial variation across the sites—both may be important depending on the
study aims.
The relationship between precision, sample size and variance is shown in
Figure 2.3. This shows that precision rapidly increases with increasing sample size
and that it does so more rapidly where there is little variance in counts between
sampling units. As a good rule of thumb, the width of the confidence intervals is
related to the number of sampling units N, as N0.5. With smaller sample sizes,
great increases in precision can be achieved for relatively small increases in sample
size. However, as sample sizes increase, so the additional precision gained declines,
and when sample sizes become very large, we gain little in precision, even for very
large increases in sample size. From a practical perspective, this tells us that if we
wish to increase precision we need to take a larger sample of sites, but beyond a
certain point, which we could think of as the optimum sample size, this produces
diminishing returns. We can use pilot data to make an informed decision about
the optimum sample size but, of course, there are often other more practical
considerations (e.g. individuals and time available for fieldwork, survey time
required within each plot, or the terrain), and the ultimate decision about sample
size will be based as much upon these as on the theory.
The other element to influence precision is the variation in counts between
sampling units. If a bird is widely and evenly distributed, occurring in roughly
Introduction| 25HighLowLowPrecisionLow varianceHigh varianceSample sizeFig. 2.3The relationship between precision and sample size with high and low
variance. Larger samples sizes deliver greater precision, but as sample sizes increase
the benefit declines.