| The evaluation of reference service requires a multi-level research design that can account for the effects of intra-class correlation that arise as the result of gathering multiple observations from each library and librarian participating in the study. Few studies have ever tested for these effects that tend to diminish the value of the standard error, misleading the investigator into believing that a relationship between variables has been discovered which does not necessarily exist. Although many previous studies have focused on a single outcome variable, this study attempted to measure the influence of seventeen predictors on three outcomes: the accuracy of the information received by the library user, the utility of the information received by the user, and the inquirer's satisfaction with the reference process. Evidence suggested that little variance in the outcome variables was explained as differences between librarians or between libraries. The potential danger of intra-class correlation was low. Variables that predicted a significant effect on the outcome measures include the extent to which the librarian followed the RUSA Behavioral Guidelines, the difficulty of the query, the user's level of education, the user's familiarity with using the library, and the level of service provided by the library. Variables that did not predict any significant effect on the outcome measures include the librarian's experience and education as well as the size of the library collection. |
