SELECTION OF SCI-TECH BOOKS: A STUDY UTILIZING A BINARY CHOICE REPEATED MEASURES MODEL, A SIMPLE CROSSOVER DESIGN, AND A FORCED CHOICE RANKING OF TEN SELECTION PRINCIPLES

Sixty-six unrelated librarians, randomly assigned to two groups, participated in an experiment requiring select/reject decisions for two sets of twenty science and technology books. A simple crossover or changeover design was used to test a hypothesis that participation in a regional resource sharing network, promising facilitated document delivery and an online circulation system for network members, would result in constrained or conservative buying. The hypothesis was supported in the case of one set of books only. For this set of books, the null hypothesis of equality of means for the total cost of books selected per librarian was rejected since a Z value of the 1.82 (p = 0.0344) was obtained. Additionally, the test statistic of the Wilcoxon matched pairs signed ranks test, T(,(-)) = 122, exceeded the critical value, allowing a rejection of the null hypothesis of equal distribution of ranks about a mean of zero for differences in rate of selection per book under the conditions of a facilitated regional resource sharing network and traditional interlibrary loan.;The favorability of book reviews was significantly correlated with rate of selection for one set of books only. The list price of books was significantly correlated with rate of selection for the other set of books only.;Librarians were also asked to rank ten representative selection principles in order of importance to their book selection decisions. Agreement in importance ranking was shown to be significant under the Friedman model (X('2)(,r) = 216). The selection principle requiring selection based on library objectives was ranked most important, but not significantly more important than selection based on community and user interest.;Probability of agreement among librarians was determined using the G-index, proposed by J. W. Holley and G. A. Linert, with respect to the binary select/reject decisions on the books and values of 0.2 - 0.3 were obtained. Probability of agreement in importance ranking of selection principles was determined using Kendall's W and values of 0.3 - 0.4 were obtained.;The binary, repeated measures model of W. G. Cochran was used to test the hypothesis that book selection by unrelated librarians is systematic rather than chance-like. The hypothesis was supported in four replications with Q values of 151.6, 158.8, 177.0, and 149.7.