The Fussler sampling technique

As indicated in Bookstein’s introduction [l] sampling is one of the most heavily used methodologies in library research. Moreover, sampling in libraries is not confined to library research alone: also other disciplines occasionally use this technique for research purposes. In this paper we will give a new and complete proof that Fussler’s sampling technique is always better than sampling by length. Bookstein emphasizes the fact that selecting the items comprising the sample is a task involving more difficulties than usually acknowledged in the literature. Moreover, rather subtle oversights can result in samples that seriously diverge in their properties from those of the population of interest. Further, the existence of distortions does not become apparent unless the research is repeated using different sampling techniques. We will compare Fussler’s method with simple random sampling, where one picks a book at random and with sampling by length where one takes random units of length rather than picking at random the ordinal position of a book. This last method however produces a bias in favor of thicker books. The solution proposed by Fussler [2] is the following: first, use sampling by length, but then select for the sample the nth book (n stays fixed) behind the one located at the randomly selected site. This method has the same advantages as sampling by length (one does not have to count hundreds of books, using large random tables), and as we will show, is at least as good and usually better. Bookstein [l] and Fussler [2] described this method for the case of sampling in card files. Here we will consider the