| Summary: | Studies utilizing content analysis have appeared with increasing frequency in literature of the social sciences over the past thirty years. During this period this technique has been employed in a wide variety of sociological researches in such areas of interest as culture and cultural change, race and ethnic relations, the sociology of war and social psychology. It is the purpose of this paper, however, to suggest that a major limitation is frequently encountered when sampling content in this manner, and must be taken into consideration both in deciding whether to use this design in a particular research situation and in the formulation of probability statements about the results. It can be shown from the theory of cluster sampling that where the elements within a duster are positively correlated, the variance of this design will be greater than would be obtained if the elements were selected according to the simple random design. In fact, relatively small positive intraclass correlations will have an appreciable effect on the variance, particularly where the size of the cluster is large.
|
|---|