ABSTRACT
Suppose that you wish to compare the size of the ant populations in two different regions of regenerated forest that have been treated in different ways. In each region, you have a number of areas available. The standard method of comparison would be to define small sample areas in each region, count the number of ants in those areas, calculate the mean and variance of the sample of counts for each region in turn, and then do a t-test to compare the means. If you were thinking carefully about the data, you might transform the counts to try to satisfy the assumption that the counts come from a normal distribution. Alternatively, if you were seriously worried about the assumptions required for a t-test, you might decide not to count the ants in each sample area, but instead simply categorize each area as sparse or abundant with respect to the ant population being considered. Clearly, by using such a crude form of assessment you are discarding information, but the actual assessment will be much quicker so that you may be able to assess many more areas from each region, and thus improve the total information.
