A collection decision model using academic health science library serials

The history of discussions on academic library serial collection management in modern times can be said to begin in the 1960s. Key variables affecting selection/deselection have emerged but without the framework of a coherent and accepted theoretical model. The advent and rapid development of a new technology of online periodicals challenges and confronts library practices and even its principles. Despite such cooperative efforts as Project COUNTER, there are still difficulties in unifying the definition of measurement units resulting in conflicting interpretations in data analysis and confusion over the meaning of these statistics.; Two methodological approaches are used to muster evidence for this investigation: statistical analyses for model construction using multiple sources of empirical data, and a survey of medical serials' librarians to offer confirmatory perspectives and to triangulate the constructed models. For the statistical analyses, serials' characteristics data were taken from three bibliographical databases in the areas of Medical and Health Sciences. This data was then augmented to include subscription information and local library environment information from an academic health science library. Factor analysis, logistic regression, and structural equation modeling were used to construct theoretical and empirical models using 14 independent variables and five factors with the serials titles' current holdings status as the dependent variable. The models are statistically significant and explain substantial variability of the dependent variable with effect sizes above 90%, indicating a robust explanation of the holdings status of the serial by a library. Importantly, the model selected is parsimonious and yet has strong predictive power when indicating the likelihood that a candidate serials title will or will not be held by the library.; Confirmatory survey data from 35 collection management librarians at moderate to large medical libraries across the United States produced results which support the statistical models and provide reasonable explanations for the components in these models. Applications of the model in practice are suggested since this methodology can be used by other libraries to assess serials holding for including and excluding titles. Expansions of this study with multiple medical libraries and with other subject areas are suggested for further study.