Crist et al. (2003) developed a randomization technique to assess the ecological nonrandom process affecting local species richness. In this method, samples were set hierarchically in regional scale, and local species richness (S_obs) are evaluated. And, all individuals in the samples to be compared are pooled and randomly reassigned to the samples, recomputing species richness expected from only random process (S_exp). They suggest (1) whether nonrandom process affects species richness or not can be evaluated by checking significance of differences between S_obs and S_exp, and (2) intensity of nonrandom process can be evaluated by S_obs/S_exp.

However the deviation of S_obs from S_exp may also depend on relative abundance distribution (RAD) in regional community and sampling effort per sample. To explain how this deviation changes with these two factors, assuming two level of hierarchy, we created a number of hypothetical communities with various combinations of intensity of nonrandom process, regional RAD and sampling effort. And then, S_obs and S_exp were computed from each hypothetical community according to the method in Crist et al. (2003).

Our results demonstrated that if sampling effort ranged from 10 to 1000 individuals per sample, significance and intensity index could reflect accurately nonrandom process on local species richness.