Geoscience Reference
In-Depth Information
So far, the discussion has been about adaptive cluster sampling for which
the initial survey is a simple random sample. Adaptive cluster sampling can
also be used with systematic sampling (Thompson, 1991a), stratified sam-
pling (Brown, 1999; Thompson, 1991b) and two-stage sampling (Salehi and
Seber, 1997). For example, in Figure  3.2 the blue-winged teal population is
sampled using stratified adaptive cluster sampling. The site is delineated
into two strata, and the initial sample of 10 quadrats allocated proportional
to the relative stratum size. In the figure, the initial sample size of the top
stratum is four quadrats, and the stratum below has an initial sample size
of 6 quadrats. The final sample size in the lower stratum stays the same, and
in the top stratum adaptive selection of additional quadrats resulted in a
within-stratum final sample size of 10 quadrats.
One final point on the design of adaptive cluster sampling is a concern often
raised with adaptive designs, namely, that the size of the final sample is not
known prior to sampling, and this makes planning the fieldwork difficult.
60
60
1
122
114
3
1
122114
3
7144
6339
14
7144
6339
14
103
150
6
103
150
6
10
10
2
2
2
2
3
3
12
12
2
2
2
2
4
4
5
20
5
20
3
3
FIGURE 3.2
The population of blue-winged teal ducks (see Figure 3.1) sampled with stratified adaptive clus-
ter sampling. The left side shows an initial sample of 10 quadrats with 4 in the top stratum and 6
in the lower stratum. The right side shows the final sample. One network larger than one quadrat
in size is selected in the final sample (the network has seven quadrats in it). The condition to trig-
ger adaptive selection was y i ≥ 1. A neighborhood was defined as the surrounding four quadrats.
Search WWH ::




Custom Search