Three types of ecological processes are responsible for variances being of different
sizes among samples. Counts of randomly dispersed animals or plants in
quadrats, cores or other sampling units will be distributed approximately as Poisson
distributions. This sort of pattern occurs where there are no ecological processes
operating to cause the individuals to aggregate or to spread out, so that they are
not scattered at random. In a Poisson distribution, the variance equals the mean.
Therefore, when comparing samples in which means differ, it is also inevitable that
variances will differ (Fig. 1.9a). Comparing samples with very large differences
among means is likely to give significant values for heterogeneity of variances
(e.g. using Cochran’s test). In such situations, where the variance: mean plot is
roughly linear, the data can usually be transformed using √X + 1 to eliminate
these problems