Two Models for Predicting Subject Circulation: A Contribution to the Allocation Problem

Abstract

It is shown that different sets of variables can account for nearly equal amounts of variance when predicting book circulation by subject. Proportion of variance (R2) for two basic models containing various combinations of 21 variables were tested. The first, called the shelf list model, treated the number of library books, whose subjects match those of academic departments, as a control variable. As such, it accounted for 6Wo of the variance, thus entering the equation first. With this model, four separate significant sets emerged, each accounting for approximately thb same amount of variance (10%). They were (1) number of faculty, hardhoft, masters enrollments; (2) masters enrollments, upper-level majors, hardhoft; (3) credit hours X total enrollments, hardhoft; (41 tipper-level majors, total majors, enrollments lower. In the second, shelf list was constrained by defining the dependent variable as the proportion of shelf list circulated. With this model, three separate significant sets emerged. These were (5) masters enrollments and hardhoft (20% variance); (6) hard-soft and Ph.D. program (16%); (7) upper-level majors, and credit hours × total majors (15%). In each of the tests, no other variables were significant. Of the 21 variables, 10 did not appear in any of the 7 sets. Any of the seven sets could be used in an allocation formula. The deciding criterion for choosing a set may depend on the convenience of collecting data for each of the variables in the set.