5. Sampling: Practical 5a
Keypoints
Keypoints
-
Simple random sampling, stratified sampling and cluster sampling are all valid techniques to take a sample from a population. Sometimes only one of these techniques is feasible in practice.
-
Which of these different method leads to the most precise (=small bias) and accurate (=small spread) estimate of population parameters like the mean with the least effort depends on the population structure.
- Often a simple random sample is the technique that gives the best results.
P.S.
The BCI dataset has been used to study several questions in community ecology, an example of such a study is the one by Condit et al. (2012), which aimed at partially validating a theoretical model ('the neutral theory of community ecology'): the number of new species (per new tree seedling) that would need to appear in the plot over time to maintain its biodiversity.
If you are interested, have a look at the paper, you can find it here: https://doi.org/10.1371/journal.pone.0049826. It is for example quite interesting to read through the Methods-subsection 'Plot census' to get an insight into the difficulties to identify some of the rare species. The lesson to be learned here: real-world observations that may seem trivial at first sight, often turn out to be more complex.
Also, through this example study it becomes clear that the BCI forest plot had to be surveyed completely (rather than taking a small sample from the area) to answer the research questions. After all, these focused especially on changes in rare species which would easily be missed when sampling ... in fact, the variation due to sampling would be larger than the value to be estimated (the change in species composition among consecutive surveys).
Condit, Chisholm, and Hubbell (2012) Thirty years of forest census at Barro Colorado and the Importance of Immigration in maintaining diversity. PLoS ONE, 7:e49826.
