SAMPLE SIZE IN DESCRIPTIVE STUDIES OF COMMUNITIES OF SESSILE AND SEMI-SESSILE BENTHIC ORGANISMS

Abstract

No one can deny the great value of adequate sample sizes to give accurate pictures of ecosystem structures and processes. The lack of a synthesis in that subject has led me to review this issue for researchers in Ecology. Initially I enumerate the main conditions required to obtain a satisfactory sample, including previous determination of study site heterogeneity and of the best shape, size and number of units (S.U.'s) to be sampled. Following an analysis of S.U. size related aspects, I review two ways for determining the ideal S.U. number in the entire sample: (1) using theoretical statistical models and their properties, and. on the other hand, (2) adopting spectral strategies. In the first case I present several techniques and algorithms used when the variables are normally distributed or when fitting Poisson or negative binomial distributions. In addition 1 include considerations on the premises involved and how to achieve them, at least in part, when the data do not fulfill such conditions. In the second case 1 consider some techniques to increase sampling area and to examine the behavior of curves obtained by plotting descriptor values against area or sample size, fitted to some common mathematical model (e.g. power, exponential or Michaelis-Menten functions). Finally, 1 discuss some strategies to define cost-benefit relationships as expressed by Molinier and Calleja points, CAIN & CASTRO criterion and auto-correlation.