期刊名称:Evidence Based Library and Information Practice
印刷版ISSN:1715-720X
电子版ISSN:1715-720X
出版年度:2016
卷号:11
期号:1
页码:4-22
DOI:10.18438/B8F918
语种:English
出版社:University Of Alberta
摘要:Abstract Objective – Measures of trends in Iowa State University library website visits per student/faculty/staff headcount show decreased use. Analysis was conducted to test for a relationship between this decrease and decreasing graduate/undergraduate enrollment ratios and decreasing visits to a popular digital collection. The purpose was to measure the influence of these factors and to produce an adjusted measure of trend which accounts for these factors. Methods – Website transaction log data and enrollment data were modelled with Box and Jenkins time series analysis methods (regression with ARMA errors). Results – A declining graduate to undergraduate enrollment ratio at Iowa State University explained 23% of the innovation variance of library website visits per headcount over the study period, while visits to a popular digital collection also declined, explaining 34% of the innovation variance. Rolling windows analysis showed that the effect of the graduate/undergraduate ratio increased over the study period, while the effect of digital collection visits decreased. In addition, estimates of website usage by graduate students and undergraduates, after accounting for other factors, matched estimates from a survey. Conclusion – A rolling windows metric of mean change adjusted for changes in demographics and other factors allows for a fairer comparison of year-to-year website usage, while also measuring the change in influence of these factors. Adjusting for these influences provides a baseline for studying the effect of interventions, such as website design changes. Box-Jenkins methods of analysis for time series data can provide a more accurate measure than ordinary regression, demonstrated by estimating undergraduate and graduate website usage to corroborate survey data. While overall website usage is decreasing, it is not clear it is decreasing for all groups. Inferences were made about demographic groups with data that is not tied to individuals, thus alleviating privacy concerns.
其他摘要:Objective – Measures of trends in Iowa State University library website visits per student/faculty/staff headcount show decreased use. Analysis was conducted to test for a relationship between this decrease and decreasing graduate/undergraduate enrollment ratios and decreasing visits to a popular digital collection. The purpose was to measure the influence of these factors and to produce an adjusted measure of trend which accounts for these factors. Methods – Website transaction log data and enrollment data were modelled with Box and Jenkins time series analysis methods (regression with ARMA errors). Results – A declining graduate to undergraduate enrollment ratio at Iowa State University explained 23% of the innovation variance of library website visits per headcount over the study period, while visits to a popular digital collection also declined, explaining 34% of the innovation variance. Rolling windows analysis showed that the effect of the graduate/undergraduate ratio increased over the study period, while the effect of digital collection visits decreased. In addition, estimates of website usage by graduate students and undergraduates, after accounting for other factors, matched estimates from a survey. Conclusion – A rolling windows metric of mean change adjusted for changes in demographics and other factors allows for a fairer comparison of year-to-year website usage, while also measuring the change in influence of these factors. Adjusting for these influences provides a baseline for studying the effect of interventions, such as website design changes. Box-Jenkins methods of analysis for time series data can provide a more accurate measure than ordinary regression, demonstrated by estimating undergraduate and graduate website usage to corroborate survey data. While overall website usage is decreasing, it is not clear it is decreasing for all groups. Inferences were made about demographic groups with data that is not tied to individuals, thus alleviating privacy concerns.