An empirical technique for determining sample size during ongoing, non‐probability, haphazard sampling is described in the present paper. While there are many techniques for determining sample size a priori, all of them require a knowledge of the population parameters (or at least their variance). Moreover, these a priori methods are based on assumptions of probability sampling such as simple random, stratified random, cluster, and so on. It is common in psychological, biological, social, environmental and medical research, however, to employ non‐probability samples of unknown representativeness and with virtually no knowledge of the parameters in question. Thus a priori techniques are quite impractical and limited in their usefulness. The procedure described in the present paper is an empirical method which does not require assumptions of probability sampling. The method involves determining the point of convergence of Sums of Squares and Cross Products (SSCP) matrices in sequential sampling using Wilk's lambda as a criterion with Rao's approximate F as a test statistic. When the SSCP matrices of two sample sequences converge at the α = .10 level of significance, the sample size is determined to be adequate as the estimators have stabilized. Further sampling would produce redundant data. An application of the present technique to a large scale study is given as a practical example.
- Sample Size