publisher colophon
abstract

This study investigated the relationship between use of online library resources and student success at a small, teaching-focused, baccalaureate college. Researchers also measured whether library users were representative of the student population. Use of online library resources was a significant predictor of semester grade point average (GPA), one-term retention, and academic standing when controlling for high school GPA, gender, status as a first-generation college student, and Pell recipient status. Library users were representative of the campus in terms of gender, ethnicity, and first-generation status. There were significant differences in the number of users of library resources by age, Pell recipient status, enrollment status, major, academic level, and semester GPA.

Introduction

Assessment is transforming higher education, and academic libraries are no exception. Resources from the federal to the local level are increasingly allocated based on impact. In this environment, the ability to quantify contributions to student success gives campus units an advantage in the decision-making process and the allocation of funds. Traditionally, the problem for academic libraries has been that demonstrating this impact involves labor-intensive, stand-alone research projects that can distract from delivery of the services being measured. Data analysis projects have the potential to change that. In this paper, researchers will present the results from one semester of an ongoing data analysis project matching use of online library resources to student success data.

Purpose of the Study

This study demonstrates that libraries can assess impact with readily available data sets. Even small institutions with limited resources can progress toward demonstrating a statistically significant relationship between library use and student success. By focusing [End Page 117] on the largest available data set and collaborating with data experts on campus, the Nevada State College Library in Henderson began a data analysis project that is both sustainable and meaningful. The library achieved this by analyzing logs on EZProxy, the proxy server through which users access online library resources, to match use of online library resources, such as databases and e-books, with student outcomes, such as grade point average (GPA), retention, and academic standing.

Working with researchers in the Office of Institutional Research (IR), the subsequent analysis project evolved into five research questions. The first question was whether users of the Nevada State College Library were representative of the student population in terms of gender, ethnicity, status as a first-generation college student, age, Pell recipient status, major, academic level, enrollment status, and semester GPA. Second, were there any proportional differences between library users and nonusers regarding their one-term retention, academic standing, or rates of DFWI (receiving a grade of D, fail, withdrawal, or incomplete on a given course)? Third, were there any significant differences in semester GPA between library users and nonusers? Fourth, was there a positive relationship between semester GPA, one-term retention, good academic standing, lower DFWI rates, and number of library use sessions? Finally, if a relationship was present between library use sessions and semester GPA, one-term retention, good academic standing, or lower DFWI rates, would that relationship hold when controlling for high school GPA, first-generation status, gender, and Pell recipient status?

About Nevada State College

At Nevada State College, the application of data-driven decision-making to student success is a core strategy toward the shared campus vision of improving outcomes for underserved and first-generation college students. The college was founded in 2002 with an inaugural class of 177 students and grew to a fall 2015 enrollment of more than 3,500. The fall 2015 class represents a varied student population with more than 50 percent ethnic diversity, more than 60 percent first-generation college students, and more than 40 percent low-income students as indicated by Pell recipient status.1 By several measures, the student population at Nevada State College has been traditionally underserved by higher education. A core strategic vision of the college is that Nevada State College should not replicate the equity gaps in performance, retention, and ultimately graduation rates seen nationally with students who match these characteristics.2

Every unit on campus plays a role toward that vision. In the library, that role is to identify and close any equity gaps in use of the library and to identify and improve any library services that have a measurable positive impact on student success. That kind of work cannot be done with traditional library statistics, such as checkouts and full-text [End Page 118] article retrievals. To align with the strategic vision of the college, the library has moved away from output-based metrics and toward outcome-based assessment measures.

Literature Review

Library Use and Student Success

In 2010, the Association of College and Research Libraries (ACRL) commissioned a report on the state of assessment in academic libraries.3 Among the next steps identified in the report was the charge for libraries to "develop systems to collect data on individual user behavior, while maintaining privacy."4 In the years surrounding that report, several libraries began to pioneer projects in user data collection. The University of Wollongong in Australia was one of the first libraries to publish such work. In 2010, the University of Wollongong Library presented at the Library Assessment Conference in Baltimore about its Library Cube, a customized database and reporting system that utilized anonymous data about students who used library resources and matched it to other data about those students, such as their course grades. The purpose of the project was to demonstrate whether the use of library resources could be shown to have a significant positive impact on student success.5 Ultimately, researchers at Wollongong found a strong correlation between multiple library use measures and student success outcomes.6 In 2011, the University of Huddersfield embarked on a similar data analysis project that expanded to include several institutions in the United Kingdom.7 In addition to using data to demonstrate a relationship between library use and graduation rates,8 the Huddersfield project also analyzed such factors as library use by discipline and academic level.9 The University of Minnesota Libraries in Minneapolis also became involved in student success research with a collection of library use data from first-year students in 2011 and 2013. Researchers at Minnesota found correlations between library use and such measures as GPA and retention.10 The researchers also explored relationships between library use and self-reported measures of academic rigor and student engagement.11

This study follows the model set by the institutions cited, with a few key differences. The first is the institutional setting. Nevada State College is a small, teaching-focused institution that serves exclusively undergraduate students. The college is classified as "Baccalaureate Colleges: Diverse Fields" in the 2015 Basic Carnegie Classification of Institutions of Higher Education. At least one other institution with this classification has demonstrated a statistically significant correlation between library use and student success measures, but it utilized a sample of 75 students rather than the entire student population.12 Due to the way the data for this project were secured and anonymized, the researchers could work with a data set that encompassed the entire student enrollment for fall 2015.

Another difference between this study and previous research is the way in which library use was measured. In many other studies of this kind, several types of library use were considered, such as online resource use, print material checkouts, study room usage, and others. At Nevada State College, the data set included only use of online resources. However, the online resource usage at Nevada State College is a more comprehensive measure of library use than in most libraries because the library collection [End Page 119] is entirely digital and usage is measured for students accessing resources both on- and off-campus. The decision to study online resource use only was made to narrow the project to a reasonable scale for a small staff.

Analyzing EZProxy Logs

EZProxy is an authentication system for library resources provided by the Online Computer Library Center (OCLC). EZProxy is most commonly used as a supplement to IP (Internet Protocol) address authentication for online databases and other resources to grant authorized users access from off-campus.13 At Nevada State College, EZProxy is used to authenticate users both on- and off-campus. As a result, the log files used for assessment are more complete than at many other institutions.

EZProxy logs are an optional feature of EZProxy that track usage data. They are most commonly used for troubleshooting or security purposes, are deleted after a few weeks, and include the IP address, date and time of request, and the complete uniform resource locator (URL), which can include search terms or items being accessed. At Nevada State College, the log file is set up to include additional fields, the most significant of which is the user ID of the individual logging in. The collection of user IDs requires strong measures to maintain privacy, which will be discussed later in this article.

Not all the information in the EZProxy log files is directly relevant to assessment projects at Nevada State College, so the first step in analyzing them is to clean them. This is done in part by a Powershell script written by the systems analyst in the Office of Information Technology Services. The script parses the log files into a comma-separated format and passes the data securely to the Office of Institutional Research (IR). The Office of IR replaces the user ID with an alternate surrogate ID number not associated with the student's record and transforms the IP address into a two-factor designation of on- or off-campus. The library gets access to that anonymized data set in a SQL (Structured Query Language) database and uses SQL UPDATE queries to pare the fields down to the minimum amount of data useful for assessment—generally, the platform or database being accessed. Most of these data are gathered for collection development purposes.14 This study considered only data on whether a student logged in to EZProxy and the total number of EZProxy sessions per student. The resources used, dates and times of requests, on- or off-campus access, and amount of time using library resources were not considered.

After the library edits the data in SQL, IR staff match the library data with institutional data about the students, such as their major, academic level, and GPA, using surrogate IDs in both data sets. This process ensures that identifiers, such as names or contact information, do not remain in any of the data sets. IR staff transform the matched data set into visualizations in Tableau data visualization software, or into statistical analyses such as those shared later in this article. To further ensure privacy, students representing a group of fewer than 10 are not reported in the results.

Data and Privacy

Obviously, the process of collecting and analyzing these data raises several questions about privacy. EZProxy logs are the digital equivalent of reading history and, as such, [End Page 120] must be secured and protected to respect users' privacy. The Office of Information Technology Services encrypts the EZProxy log data, and the Office of IR anonymizes them. The former strategy is intended to secure the raw data from access by unauthorized parties, and the latter helps to mask the identity of individual users from staff not authorized for access to user data. Due to the vulnerability of raw anonymous data to reidentification attacks,15 which use surrogate data sets to reidentify masked users based on information shared about them, the Office of IR also practices aggregation. Raw data sets are not shared publicly even in their anonymous form, and reported results are clustered into groups of 10 or more users.

The log-in page for EZProxy includes a notice about cookie collection that links to a statement about the data collection project with the full Terms of Use and Privacy Policy. Nevada State College recently updated the policy to address many of the concerns in the National Information Standards Organization (NISO) Consensus Principles document on user's digital privacy, published in the fall of 2015.16 Students have the option to opt out by contacting the library. In the case of an opt out, any data that have not been anonymized will be deleted, future data associated with that student ID will be deleted in parsing, and the student will receive an anonymous log-in for future use. No student in fall 2015 elected to opt out.

At Nevada State College, the approach to privacy involves choice, protection, and constant communication with constituents through meetings, presentations, advisory boards, and publicly available information and notices. Despite this, the authors recognize that there are still shortcomings in the process that we continually work to address. This kind of approach has potential risks, but the possible benefits for improving student learning may outweigh them. The potential to help students succeed, to close equity gaps, and to improve learning outcomes is too important to ignore while waiting for a perfect solution, especially in a field that involves technology and security, where the perfect solution may always be a moving target. The risks will never dissipate completely, but by mitigating them and striving toward continuous improvement, we can set reasonable expectations for our users and minimize the risk that their data will be misused.

Methods

Context of the Study

This study took place in a comprehensive, nonresidential, four-year public state college in the Southwestern United States that is dedicated to excellence in teaching and learning. Nevada State College in Henderson offers baccalaureate degrees in more than 20 fields of study, with 43 majors and minors in the School of Education, the School of Liberal Arts and Sciences, the School of Nursing, and the Department of Business Administration. The college prepares a diverse and largely underserved student population for the state's [End Page 121] high-demand industries, including health care and education. In fall 2015, the college had a student enrollment of more than 3,500 and was classified as a baccalaureate college with diverse fields by the Carnegie Foundation for the Advancement of Teaching.

The library at Nevada State College is the first entirely digital library in the state. It provides immediate access to over 1.4 million e-books through a combination of subscriptions and demand-driven purchasing. The library also provides access to almost 2 million print volumes through interlibrary loan with partner libraries in southern Nevada; however, print circulation accounts for only 1.3 percent of total collection use. In addition to offering traditional library services, such as one-on-one or small group research consultations, librarians participate in first-year student experience courses and offer instructional design services related to information literacy, such as research assignment design. The academic faculty in the library also closely collaborate with the Office of IR to track students' usage of library online resources to gain a deeper understanding of library use and make evidence-based decisions about how to provide better services to help students succeed.

Participants

Participants of this study were students who enrolled in fall 2015. Data from a total of 3,530 students registered in fall 2015 were included. The characteristics and distributions of the students based on library user and nonuser status are presented in Tables 1 and 2. Data from 1,936 students who have used library resources were employed to address the fourth research question of this study: "Was there a positive relationship between semester GPA, one-term retention, good academic standing, lower DFWI rates, and number of library use sessions?" Since only first-time students (not transfer students) were required to submit their high school transcripts, data from 638 students who have high school GPAs available and have used library resources were utilized to address the fifth research question: "If a relationship was present between library use sessions and semester GPA, one-term retention, good academic standing, or lower DFWI rates, would that relationship hold when controlling for high school GPA, first-generation status, gender, and Pell recipient status?"

Variables of Interest

Exploratory variables involved in this study included student demographics, enrollment variables (for example, full-time or part-time enrollment status, major, and academic level), and library user or nonuser status. Student demographics were gender, firstgeneration status, ethnicity, Pell recipient status, as well as enrollment status (full-time or part-time) and high school GPA, which previous studies found to be related to college performance.17 High school GPA was acquired from the student's high school transcript and measured on a 4.0 scale. Gender in this study was coded as 0 for female and 1 for male. First-generation status in this study was defined as neither parent earned a bachelor degree and was coded as 0 for "not first-generation" and 1 for "first-generation." Pell recipient status refers to whether a student ever received a Pell Grant, the largest federal need-based financial aid program available to postsecondary education students across the United States. The Pell Grants are awarded mostly to low-income students, based [End Page 122] primarily on the student's or parents' income for the previous year.18 Therefore, we use the Pell Grant recipient status as a proxy for low-income status. Students who did not receive a Pell Grant were coded as 0, and students who have ever received a Pell Grant were coded as 1. Enrollment status was measured by academic load and was categorized as two subgroups, full-time and part-time. Part-time was coded as 0, and full-time as 1. The library usage was coded as 0 if a student did not use library resources and 1 if a student used library resources at least once in fall 2015. In addition, the number of library sessions was measured by how many times a student logged in to online library resources through EZProxy in fall 2015.

The outcome variables of this study were student semester GPA, one-term retention, academic standing, and DFWI. Semester GPA was measured on a 4.0 scale. Those students who did not have semester GPAs available due to withdrawing from all courses were excluded in the data analysis. One-term retention was fall to spring retention and was coded as 0 if a student was not retained and 1 if he or she was retained. Academic standing was measured by letter grades. Students who earned a letter grade of A, B, or C for the semester were coded as 1 for having good academic standing. Otherwise, the student was coded as 0 for lacking good academic standing. Similar coding was used for DFWI: 0 for a student receiving no DFWI grades and 1 for having at least one DFWI.

Data Analysis

Due to the continuous as well as categorical nature of the outcome variables of this study, the researchers performed parametric and nonparametric statistical analyses to address the research questions posed in the Introduction section. To address the first research question of how representative library users were of the campus population, the number and percentage of each subgroup based on library user and nonuser status were presented in Tables 1 and 2. The chi-squared or χ2 test has become one of the well-used statistical analyses in educational research since its inception. It can be applied for categorical variables and for testing goodness of fit, independence, and homogeneity.19 Therefore, the chi-squared test was used to examine the proportional differences between library users and nonusers regarding their demographics and enrollment variables, to see whether users of the Nevada State College Library were representative of the campus.

To answer the research question "Is there any significant difference regarding semester GPA between library users and nonusers?" a one-way analysis of variance (ANOVA) was conducted to compare the mean difference of semester GPA for library users and nonusers. Although another statistical test, the t-test, may be an option for such a comparison, ANOVA was selected because it allows for a robust test of equality of means.20

Due to the categorical nature of the other three outcome variables (one-term retention, academic standing, and DFWI) and one exploratory variable of library use status (user and nonuser), chi-squared tests were employed to examine the proportional differences between library users or nonusers in terms of these three outcome variables. Three chi-squared tests were performed to examine the proportional differences in one-term retention, good academic standing, and DFWI rates for library users and nonusers.

To examine the relationship between number of library use sessions (number of EZProxy sessions) and student success indicators (semester GPA, one-term retention, [End Page 123] academic standing, and DFWI), a Pearson correlation was performed to examine the strength of the relationship between number of library use sessions and semester GPA. A Spearman correlation, another test that indicates the strength of the relationship between two variables, was conducted to examine the relationships between number of library use sessions and one-term retention, good academic standing, and DFWI, respectively.

To answer the fifth research question of whether any relationship between number of library use sessions and semester GPA would hold when controlling for other variables, a method called hierarchical linear regression was used to test for the influence of different variables by adding them to the model one at a time, for the variable semester GPA on library use. Three hierarchical logistic regressions were performed for one-term retention, academic standing, and DFWI, respectively, controlling for high school GPA, gender, first-generation status, and Pell recipient status. Although other factors may be related to student outcomes, the rationale of controlling for these four variables was that these variables are also found to impact student performance in the literature.21 Most importantly, we focused our attention on the impact of library use on college success. As a result, we conducted a hierarchical regression by including library use sessions as one block of the variable, in addition to high school GPA, gender, first-generation status, and Pell recipient status as one block of control variables. These two blocks of variables were entered into the predictive equation in a hierarchical order to examine the R-squared change (adjusted R-squared change), a measure of the increase in predictive power (R squared) resulting from the inclusion of a new predictor (or block of predictors), and the significance of F value change (a measure of whether the final model significantly improves the ability to predict the outcome variable of semester GPA) with an additional block of variables introduced. For all tests, the alpha level for statistical significance was set at 0.05.

Limitations

It could be meaningful to perform an analysis on one-year retention and graduation rate. However, since the library use and student success initiative started recently and only the inaugural data were available at this time, more longitudinal data are needed to examine the long-term effects of such innovation. Therefore, we focused our examinations on semester GPA, one-term retention, academic standing, and DFWI rates, which were found to be strong and important indicators of college success based on the literature.22

For this study, we controlled for several variables that could impact library use and academic success, such as high school GPA, first-generation status, gender, and Pell recipient status. We could not control for all potential variables. In particular, it would have been useful to analyze whether enrollment status (full-time or part-time) had an impact on our results because part-time students may have less need for library resources.

It would be useful to consider an analysis based on amount of time spent in library databases rather than just number of sessions. However, EZProxy does not record a true end-of-session time. Future studies could approximate session length based on the last activity in EZProxy, but such a calculation would be an estimate only because some electronic resource platforms reference EZProxy more frequently than others, creating a longer approximated session time. [End Page 124]

Another limitation of the study is that EZProxy sessions are only one measure of library use. For future analyses, tracking research consultations with a librarian, information literacy instruction session attendance, or measures of library as space would be helpful. The library began collecting additional data points in 2017 with the hope of including them in future studies.

Finally, the correlational nature of the data prohibits the identification of any causal relationships between library use and student success indicators. As a result, the findings of this study need to be interpreted with caution.

Results

Data analyses led to the following salient findings. The demographic proportions of students on campus based on their status as library resource users and nonusers are presented in Table 1. There were 3,530 students enrolled in the fall semester of 2015. Among these students, 1,936 (54.8 percent) students have used the library's online resources in fall 2015. The mean number of library use sessions among these 1,936 students was 7.93 sessions, with a standard deviation of 9.91. The highest usage was 127 sessions, while 1,594 (45.2 percent) students showed no usage of the library's online resources.

As shown in Tables 1 and 2, there were no statistically significant proportional differences between library users and nonusers regarding their gender, ethnicity, and first-generation status, p > 0.05. However, there were significant proportional differences between library users and nonusers in terms of their age, Pell recipient status, major, academic level, enrollment status (full-time or part-time), and GPA. Overall, students of nontraditional age (age 25 and up), Pell recipients, nursing students, seniors, and students who earned a higher GPA were more likely to use library resources than their peer counterparts, p < 0.05. The reasons more of these students used library resources than their peers are beyond the scope of the paper but may be worth further examination in the future.

Second, there was a significant difference in terms of semester GPA between library users and nonusers. We conducted an ANOVA to compare the mean difference of student semester GPA. The Levene statistic for the test of the equality of variances was significant, p (the probability that the difference between the two values can be attributed to chance alone) = 0.009, which suggests a violation of the assumption of homogeneity of variances. Therefore, alternatively, Welch's F test was performed, since it is a robust test [End Page 125]

Table 1. Proportional differences between library users and nonusers by demographics * May not sum up to the total (3,530) because of missing values. † χ2 or chi-squared is a statistical test employed to compare observed data with the results expected. The larger the chi-squared value, the greater the probability that a significant proportional difference exists. ‡ p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance. § ns = no significant difference. ** Bold indicates significantly higher proportions.
Click for larger view
View full resolution
Table 1.

Proportional differences between library users and nonusers by demographics

* May not sum up to the total (3,530) because of missing values.

χ2 or chi-squared is a statistical test employed to compare observed data with the results expected. The larger the chi-squared value, the greater the probability that a significant proportional difference exists.

p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

§ ns = no significant difference.

** Bold indicates significantly higher proportions.

[End Page 126]

Table 2. Proportional differences between library users and nonusers by enrollment variables *May not sum up to the total (3,530) because of missing values. † χ2 or chi-squared is a statistical test employed to compare observed data with the results expected. The larger the chi-squared value, the greater the probability that a significant proportional difference exists. ‡ p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance. § Bold indicates significantly higher proportions.
Click for larger view
View full resolution
Table 2.

Proportional differences between library users and nonusers by enrollment variables

*May not sum up to the total (3,530) because of missing values.

χ2 or chi-squared is a statistical test employed to compare observed data with the results expected. The larger the chi-squared value, the greater the probability that a significant proportional difference exists.

p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

§ Bold indicates significantly higher proportions.

[End Page 128] of equality of means. Students who used library resources had a semester GPA higher than that of their peers who did not use library resources, as presented in Table 3. For library users, M (mean) = 3.16 and SD (standard deviation) = 0.85, while for nonusers, M = 2.99 and SD = 0.88. Welch's F(1, 2790.74) = 27.30, p < 0.01, ƞ2 = 0.01 indicated a small effect size, meaning that only a small proportion of the variance in semester GPA can be attributed to library user or nonuser status.23

Third, the results also showed significant proportional differences between library users and nonusers regarding their one-term retention rates and good academic standing. However, no significant proportional difference was found between library users and nonusers regarding their DWFI rates.

Results of a chi-squared test indicated a significant proportional difference in one-term retention rates between students who used library resources (84.9 percent retained) and that of their peers who did not use the library resources (66.0 percent retained),2 (1) = 173.43, p < 0.01, phi (φ) coefficient (a measure of association between two binary variables) = 0.222, indicating a small to medium effect size.24 The results suggest that students who used library resources were retained at a significantly higher rate than their peers who did not use the library.

Similarly, there was a significant proportional difference in good academic standing between students who used the library (89.0 percent on good academic standing) and that of their peers who did not use the library (79.1 percent on good academic standing), χ2(1) = 65.03, p < 0.01, phi (φ) coefficient = 0.137, indicating a small effect size.25 The results indicate that students who used library resources were more likely to have good academic standing than those who did not use library resources.

However, there was no significant proportional difference in DFWI rates between students who used the library (18.1 percent with a DFWI grade) and DFWI rates of their peers who did not (18.7 percent with a DFWI grade), χ2(1) = 0.186, p > 0.05, phi (φ) coefficient = 0.007, indicating a small to medium effect size.26 Although students who used the library had a slightly lower DWFI rate, the proportional difference was not statistically significant.

The results of this study also revealed that students who are retained and have good academic standing had a significantly higher number of library use sessions than those who are not retained and who lack good academic standing, p < 0.01 (see Table 4). There was no significant difference in number of library use sessions between students who have a DFWI grade than those who do not have a DWFI.

Fourth, the results of this study indicated that there were positive and statistically significant relationships between semester GPA, one-term retention, good academic standing, and library use measured by number of sessions for the library users. Results of correlational analysis revealed a significantly positive but a weak relationship between semester GPA and number of library sessions, r = 0.200, p < 0.01. Spearman correlational analysis revealed a significant and positive relationship between one-term retention and number of library sessions, Spearman's rho = 0.135, p < 0.01. Also, a significant and positive relationship between good academic standing and number of library sessions was found, Spearman's rho = 0.198, p < 0.01. There was a negative relationship between DFWI and number of library sessions for the library users. However, the relationship was not statistically significant, Spearman's rho = –0.028, p > 0.05. [End Page 129]

Table 3. Semester GPA, term retention, academic standing, and DFWI* by library user and Nonuser * DWFI = receiving a grade of D, fail, withdrawal, or incomplete. † Bold indicates statistically significant higher mean or proportions. ‡ Welch's F test is designed to test the equality of group means when the assumption of homogeneity of variance was Not met. § p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance. ** χ2 or chi-squared is a statistical test employed to compare observed data with the results expected. The larger the chi-squared value, the greater the probability that a significant proportional difference exists.
Click for larger view
View full resolution
Table 3.

Semester GPA, term retention, academic standing, and DFWI* by library user and Nonuser

* DWFI = receiving a grade of D, fail, withdrawal, or incomplete.

Bold indicates statistically significant higher mean or proportions.

Welch's F test is designed to test the equality of group means when the assumption of homogeneity of variance was Not met.

§ p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

** χ2 or chi-squared is a statistical test employed to compare observed data with the results expected. The larger the chi-squared value, the greater the probability that a significant proportional difference exists.

[End Page 130]

Table 4. Means and standard deviations* of library use sessions by one-term retention, academic standing, and DFWI† status * Standard deviation indicates how tightly the data cluster around the mean. † DWFI = receiving a grade of D, fail, withdrawal, or incomplete. ‡ F, the F-statistic, is a ratio of two sample variances. Variances are a measure of how far the data are scattered from the mean. Larger F values represent greater dispersion. § p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.
Click for larger view
View full resolution
Table 4.

Means and standard deviations* of library use sessions by one-term retention, academic standing, and DFWI† status

* Standard deviation indicates how tightly the data cluster around the mean.

DWFI = receiving a grade of D, fail, withdrawal, or incomplete.

F, the F-statistic, is a ratio of two sample variances. Variances are a measure of how far the data are scattered from the mean. Larger F values represent greater dispersion.

§ p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

Fifth, the results from linear and logistic regression analysis indicated that, after controlling for high school GPA, gender, first-generation status, and Pell recipient status, library use was still a significant predictor of semester GPA, one-term retention, and academic standing. In terms of semester GPA, with the introduction of library sessions to the model, R squared increased 3.7 percent, from 0.141 to 0.178, and adjusted R squared changed from 0.135 to 0.171, which suggests that about 3.7 percent of the total variance in semester GPA can be accounted for by addition of the library use variable to high school GPA, gender, first-generation status, and Pell recipient status, F (5, 601) = 27.37, p < 0.001. High school GPA and library use were two significant predictors of semester GPA, t(602) = 8.93, p < 0.001, and t(602) = 5.23, p < 0.001, respectively (see Table 5). The results revealed that library use measured by sessions was significantly related to semester GPA, indicating [End Page 131] that a one-unit increase in library use sessions would likely correspond with an increase of semester GPA by 0.029. Similarly, a one-unit increase in high school GPA would likely correspond with an increase of semester GPA by 0.615.

Regarding oneterm retention, a logistic regression on library use controlling for high school GPA, gender, first-generation status, and Pell recipient status was performed. The results indicated that, after holding other variables constant, number of library sessions was a significant predictor of one-term retention, Wald F(1) = 16.64, p < 0.001 (see Table 6). Wald is a way of testing the significance of explanatory variables in a statistical model. If there was a one-unit increase in library sessions, the odds of this student being retained next term would increase by 16.9 percent. High school GPA was also a significant predictor of one-term retention, Wald F(1) = 7.78, p < 0.01. If there was a one-unit increase in a student's high school GPA, the odds that this student would be retained next term would increase by 93.6 percent. If a student received a Pell Grant, the odds [End Page 132] that the student would be retained next term would increase by 74.5 percent, compared to a student who did not receive Pell funding, Wald F(1) = 5.05, p < 0.05. If a student is first-generation, the odds that the student would be retained next term would decrease by 39.3 percent compared to a non-first-generation student, Wald F(1) = 3.88, p < 0.05. Gender was not a significant predictor of one-term retention, Wald F(1) = 0.168, p > 0.05.

Table 5. Results of hierarchical linear regression* of semester GPA on library use * Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time. † B, the unstandardized coefficient, is the regression coefficient for each independent variable. ‡ Standard error measures the accuracy with which a sample represents a population. § β, the standardized coefficient, is the regression coefficient for each independent variable. ** t is a statistical test of whether the B value is significantly different from zero. †† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance. ‡‡ Bold indicates especially a statistically significant predictor of an independent variable of the dependent variable.
Click for larger view
View full resolution
Table 5.

Results of hierarchical linear regression* of semester GPA on library use

* Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time.

† B, the unstandardized coefficient, is the regression coefficient for each independent variable.

‡ Standard error measures the accuracy with which a sample represents a population.

§ β, the standardized coefficient, is the regression coefficient for each independent variable.

** t is a statistical test of whether the B value is significantly different from zero.

†† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

‡‡ Bold indicates especially a statistically significant predictor of an independent variable of the dependent variable.

Table 6. Results of hierarchical logistic regression* of one-term retention on library use * Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time. † B lists the partial logistic regression coefficients for each independent variable. ‡ S.E. or standard error measures the accuracy with which a sample represents a population § Wald is a way of testing the significance of explanatory variables in a statistical model. ** df or degree of freedom is the number of values in the study that are free to vary. †† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance. ‡‡ Exp(B) list the odds ratios, which are used to assess the isolated impact of each independent variable. §§ Bold indicates statistically significant results.
Click for larger view
View full resolution
Table 6.

Results of hierarchical logistic regression* of one-term retention on library use

* Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time.

B lists the partial logistic regression coefficients for each independent variable.

S.E. or standard error measures the accuracy with which a sample represents a population

§ Wald is a way of testing the significance of explanatory variables in a statistical model.

** df or degree of freedom is the number of values in the study that are free to vary.

†† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

‡‡ Exp(B) list the odds ratios, which are used to assess the isolated impact of each independent variable.

§§ Bold indicates statistically significant results.

Regarding academic standing, a logistic regression was conducted on library use controlling for high school GPA, gender, first-generation status, and Pell recipient status. The results indicated that, after controlling for other variables, number of library sessions was a significant predictor of good academic standing, Wald F(1) = 13.84, p < 0.001 (see Table 7). If there was a one-unit increase in library sessions, the odds of the student having good academic standing would increase by 13.2 percent. High school GPA was also a significant predictor of academic standing, Wald F(1) = 39.91, p < 0.001. If there was a one unit increase in a student's high school GPA, the odds of that student having good academic standing would increase by 376 percent. Gender, first-generation status, and Pell recipient status were not significant predictors of academic standing, Wald F(1) = 0.86, 1.65, and 0.405, p > 0.05, respectively. [End Page 133]

Table 7. Results of hierarchical logistic regression* of academic standing on library use * Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time. † B lists the partial logistic regression coefficients for each independent variable. ‡ S.E. or standard error measures the accuracy with which a sample represents a population § Wald is a way of testing the significance of explanatory variables in a statistical model. ** df or degree of freedom is the number of values in the study that are free to vary. †† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance. ‡‡ Exp(B) list the odds ratios, which are used to assess the isolated impact of each independent variable. §§ Bold indicates statistically significant results.
Click for larger view
View full resolution
Table 7.

Results of hierarchical logistic regression* of academic standing on library use

* Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time.

B lists the partial logistic regression coefficients for each independent variable.

S.E. or standard error measures the accuracy with which a sample represents a population

§ Wald is a way of testing the significance of explanatory variables in a statistical model.

** df or degree of freedom is the number of values in the study that are free to vary.

†† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

‡‡ Exp(B) list the odds ratios, which are used to assess the isolated impact of each independent variable.

§§ Bold indicates statistically significant results.

None of the predictors was significantly related to student DFWI rates, p > 0.05 (see Table 8). However, there is a tendency of decreasing of DFWI rate if students use more library resources (see Table 8).

Discussion

The researchers were encouraged to find no significant proportional differences in number of library users by gender, ethnicity, or first-generation status. Our findings on the demographic characteristics of library users versus nonusers contradict some studies in the recent literature. For example, the University of Huddersfield found some statistically significant differences in library use by age, gender, and ethnicity with one cohort of students graduating in 2011. However, the effect sizes were small, and there was no difference found for hours logged-in to online library resources by ethnicity.27 Similarly, researchers at Indiana University Kokomo observed slightly higher percentages [End Page 134] of library users among Asian students, women, and students aged 29 and up.28 Our study found similar patterns with gender, ethnicity, and first-generation status, but the differences were not statistically significant. Future research into equity gaps in library use should consider the number of use sessions and other types of library use beyond online resource usage, but that was not the focus of this study.

Table 8. Results of hierarchical logistic regression of DFWI* on library use * Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time. † B lists the partial logistic regression coefficients for each independent variable. ‡ S.E. or standard error measures the accuracy with which a sample represents a population § Wald is a way of testing the significance of explanatory variables in a statistical model. ** df or degree of freedom is the number of values in the study that are free to vary. †† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance. ‡‡ Exp(B) list the odds ratios, which are used to assess the isolated impact of each independent variable.
Click for larger view
View full resolution
Table 8.

Results of hierarchical logistic regression of DFWI* on library use

* Hierarchical linear regression is a statistical technique that tests for the influence of different variables by adding them to the model one at a time.

B lists the partial logistic regression coefficients for each independent variable.

S.E. or standard error measures the accuracy with which a sample represents a population

§ Wald is a way of testing the significance of explanatory variables in a statistical model.

** df or degree of freedom is the number of values in the study that are free to vary.

†† p is an estimate of the probability that the result occurred by statistical accident. A low level of p indicates a high level of statistical significance.

‡‡ Exp(B) list the odds ratios, which are used to assess the isolated impact of each independent variable.

Interestingly, students characterized as low-income due to receiving Pell Grant funds were proportionally higher users of library resources than students who were not. This seemingly contradicts a previous study from the University of Minnesota that observed lower instances of library use for low-income students in a variety of areas. However, the University of Minnesota examined several indicators of socioeconomic status, including student-reported data about their intent to seek employment, concerns about paying [End Page 135] for college, and financial aid distributions. Based on Pell recipient status alone, Krista Soria, Shane Nackerud, and Kate Peterson found no statistically significant relationship to e-book usage, journal, or database usage, which are the closest approximations to the EZProxy log sessions used in this study. Soria, Nackerud, and Peterson also utilized first-generation status as an indicator of socioeconomic status.29 Our study observed no statistically significant differences based on this variable. In the future, it may be useful to look at longitudinal data beyond one-term retention rates and to analyze additional variables, such as percentage of unmet financial need, to determine if the pattern of higher number of library users still holds.

In terms of the academic level and major of library users, findings from this study resemble findings from previous research.30 There seems to be a pattern of an increasing proportion of library users from the freshman through senior level, as demonstrated by the percentage of users by academic level in Table 2. However, the difference was only statistically significant for seniors. In terms of academic discipline, nursing students are the most likely to use library resources, while students from the physical and life sciences, business, and education disciplines are underrepresented among library users. The reasons for this discrepancy are unclear from the existing data set, but librarians at Nevada State College have already begun to use the information as a marketing tool with some academic departments. In future years, these data can serve as a benchmarking tool to assess those efforts.

This study confirms previous research that found a correlation between library use and student success measures. Some previous studies have found a strong correlation between library use and student success measures such as GPA and retention rates.31 Our study found weak correlations of library use sessions with one-term retention, academic standing, and semester GPA. We found no significant relationship with DFWI rates. There could be several reasons for the disparity. First, previous studies have not measured academic standing or DFWI rates. They have also been conducted mostly in a research university setting, rather than in a teaching-focused college library. Nevada State College probably has relatively fewer research assignments as part of the curriculum, which could impact both use levels and outcomes. Researchers at Nevada State College have begun to explore if a relationship to student success is stronger when analyzing research assignments directly, rather than aggregated measures of success such as GPA.32 It would also be useful to know if future research at other small teaching-focused institutions produces similar results.

Another potential reason for the discrepancy in strength of the correlation is that this project analyzed only use of online library resources as measured by EZProxy sessions. Despite the Nevada State College Library being a digital library that authenticates users on- and off-campus, this data set is still incomplete when considering the breadth of services and resources available to students. Previous studies have looked at additional variables, such as print book checkouts, information literacy instruction, and reference [End Page 136] transactions.33 The additional data points may be a significant contributing factor to the strength of the correlations. In the academic year 2016–2017, librarians at Nevada State College collected additional data points about library use to determine if adding these elements improves the strength of the correlation and if any specific library services or resources can be shown to have a greater relationship to student success than others.

One unique finding of this study was that library use was still a significant predictor of semester GPA, one-term retention, and academic standing after controlling for high school GPA, gender, first-generation status, and Pell recipient status. These findings are encouraging because they are more compelling than a single correlational analysis. The multiple regression used to demonstrate this finding helps to eliminate some of the other possible explanations for the relationship and shows with more confidence that academic library use plays a key role in student success.

Conclusion

The findings not only allow us to see a positive connection between library use and student success but also extend our understanding of the role an academic library plays in student success at a teaching college. The diverse student population and the digital library setting also set this research apart from previous studies. While the correlation between library use and student success is statistically significant, it is not substantial in magnitude. However, so many factors contribute to outcomes such as GPA and retention that any statistically significant impact in this area bears further investigation. Even the small effect sizes demonstrated in this study garnered positive recognition from campus administrators and faculty at Nevada State College and motivated campus leaders to extend library services to more students and to encourage students to use more library resources.

Researchers at Nevada State College plan to continue to monitor the longitudinal effects of the data presented and determine if there is any statistically significant relationship to six-year graduation rates. Researchers will also analyze interaction effects with other services on campus such as the Writing Center and tutoring to determine if any combination of these services can be shown to improve outcomes for students with a greater effect size than any one of the services in isolation. There is also interest in further exploring connections to positive student outcomes that are more directly impacted by library resources and services, such as research assignment grades.

Tiffany LeMaistre

Tiffany LeMaistre is the electronic resources and discovery librarian at Nevada State College in Henderson; she may be reached by e-mail at: Tiffany.Lemaistre@nsc.edu.

Qingmin Shi

Qingmin Shi is a quantitative analyst at Nevada State College in Henderson; she may be reached by e-mail at: Qingmin.Shi@nsc.edu.

Sandip Thanki

Sandip Thanki is the director of institutional research at Nevada State College in Henderson; he may be reached by e-mail at: Sandip.Thanki@nsc.edu.

Acknowledgments

The authors would like to acknowledge some of their colleagues at Nevada State College, without whom this work would not have been possible. In particular, Kathryn Mulvey, Janice Le-Nguyen, Mick Haney, and Nathaniel King were all instrumental to the data collection, analysis, and support of this research. [End Page 137]

Notes

1. Nevada State College, "Institutional Data," 2016, http://nsc.edu/institutional-research/institutional-data.aspx.

2. Nevada State College, "2015–2020 Academic Strategic Plan: Opportunity, Enrichment, Impact," 2015, https://nsc.edu/Files/provost/pdfs/NSC_2015-2020_Strategic_Plan.pdf.

3. Megan Oakleaf, The Value of Academic Libraries: A Comprehensive Research Review and Report (Chicago: Association of College and Research Libraries [ACRL], 2010), http://www.ala.org/acrl/sites/ala.org.acrl/files/content/issues/value/val_report.pdf.

4. Ibid., 12.

5. Margie Jantti and Brian Cox, "Measuring the Value of Library Resources and Student Academic Performance through Relational Datasets," in Steve Hiller, Kristina Justh, Martha Kyrillidou, and Jim Self, eds., Proceedings of the 2010 Library Assessment Conference: Building Effective, Sustainable, Practical Assessment (Washington, DC: Association of Research Libraries, 2011), 525–32, http://libraryassessment.org/bm~doc/proceedings-lac-2010.pdf.

6. Brian Cox and Margie Jantti, "Capturing Business Intelligence Required for Targeted Marketing, Demonstrating Value, and Driving Process Improvement," Library & Information Science Research 34, 4 (2012): 308–16, doi:10.1016/j.lisr.2012.06.002.

7. Graham Stone, Bryony Ramsden, and Dave Pattern, "Library Impact Data Project Toolkit," University of Huddersfield, 2011, http://eprints.hud.ac.uk/11571/1/Toolkit_final.pdf.

8. Graham Stone and Bryony Ramsden, "Library Impact Data Project: Looking for the Link between Library Usage and Student Attainment," College & Research Libraries 74, 6 (2013): 546–59, doi:10.5860/crl12-406.

9. Ellen Collins and Graham Stone, "Understanding Patterns of Library Use among Undergraduate Students from Different Disciplines," Evidence Based Library and Information Practice 9, 3 (2014): 51–67, http://ejournals.library.ualberta.ca/index.php/EBLIP/article/view/21326.

10. Krista M. Soria, Jan Fransen, and Shane Nackerud, "Library Use and Undergraduate Student Outcomes: New Evidence for Students' Retention and Academic Success," portal: Libraries and the Academy 13, 2 (2013): 147–64, http://conservancy.umn.edu/handle/11299/143312.

11. Krista M. Soria, Jan Fransen, and Shane Nackerud, "Beyond Books: The Extended Academic Benefits of Library Use for First-Year College Students," College & Research Libraries 78, 1 (2017), http://crl.acrl.org/content/early/2016/01/25/crl16-844.

12. Angie Thorpe, Ria Lukes, Diane J. Bever, and Yan He, "The Impact of the Academic Library on Student Success: Connecting the Dots," portal: Libraries and the Academy 16, 2 (2016): 373–92, doi:10.1353/pla.2016.0027.

13. Online Computer Library Center (OCLC), "Log File Analysis," in OCLC Support and Training, 2015, http://www.oclc.org/support/services/ezproxy/documentation/loganalysis.en.html.

14. Tiffany LeMaistre, "Cost Per User: Analyzing EZProxy Logs for Assessment," in Proceedings of the Charleston Library Conference, Charleston, SC, 2015, doi:10.5703/1288284316291.

15. Gretchen McCord, What You Need to Know about Privacy Law: A Guide for Librarians and Educators (Santa Barbara, CA: Libraries Unlimited, 2013); Paul Ohm, "Broken Promises of Privacy: Responding to the Surprising Failure of Anonymization," UCLA Law Review 57 (2010): 1701, University of Colorado Law Legal Studies Research Paper No. 9-12, available [End Page 138] at SSRN (Social Science Research Network), http://papers.ssrn.com/abstract=1450006; Aditi Ramachandran, Lisa Singh, Edward Porter, and Frank Nagle, "Exploring Re-Identification Risks in Public Domains." United States Census Bureau, Research Report Series, 2012, https://www.census.gov/srd/CDAR/rrs2012-13_Exploring_Re-ident_Risks.pdf.

16. National Information Standards Organization (NISO), "NISO Consensus Principles on User's Digital Privacy in Library, Publisher, and Software-Provider Systems (NISO Privacy Principles)," 2015, http://www.niso.org/apps/group_public/download.php/16064/NISO%20Privacy%20Principles.pdf.

17. Clifford Adelman, The Toolbox Revisited: Paths to Degree Completion from High School through College (Washington, DC: U.S. Department of Education, 2006), http://www.ed.gov/rschstat/research/pubs/toolboxrevisit/index.html; Paul Attewell, Scott Heil, and Liza Reisel, "What Is Academic Momentum? And Does It Matter?" Educational Evaluation and Policy Analysis 34, 1 (2012): 27–44, doi:10.3102/0162373711421958; Erika Beck, Tony Scinta, Robin Cresiski, Sandip Thanki, and Qingmin Shi, "Closing the Equity Gap: Student Support Services Matter," presentation at American Educational Research Association, Washington, DC, April 10, 2016, http://www.aera.net/Publications/Online-Paper-Repository/AERA-Online-Paper-Repository/Owner/976752.

18. Christina Chang Wei and Laura Horn, "Persistence and Attainment of Beginning Students with Pell Grants" (NCES 2002-169), 2002, U.S. Department of Education, Washington, DC, National Center for Education Statistics, http://nces.ed.gov/pubs2002/2002169.pdf.

19. Todd Michael Franke, Timothy Ho, and Christina A. Christie, "The Chi-Square Test Often Used and More Often Misinterpreted," American Journal of Evaluation 33, 3 (2012): 448–58, doi:10.1177/1098214011426594.

20. Gwowen Shieh and Show-Li Jan, "Determining Sample Size with a Given Range of Mean Effects in One-Way Heteroscedastic Analysis of Variance," Journal of Experimental Education 81, 3 (2013): 281–94, doi:10.1080/00220973.2012.731020.

21. Adelman, The Toolbox Revisited; Attewell, Heil, and Reisel, "What Is Academic Momentum?"; Beck, Scinta, Cresiski, Thanki, and Shi. "Closing the Equity Gap."

22. Brice W. Harris, "Looking Inward: Building a Culture for Student Success," Community College Journal of Research and Practice 22, 4 (1998): 401–18, doi:10.1080/1066892980220407; George D. Kuh, Jillian Kinzie, Jennifer A. Buckley, Brian K. Bridges, and John C. Hayek, What Matters to Student Success: A Review of the Literature (Washington, DC: National Postsecondary Education Cooperative, 2006).

23. Jacob Cohen, Statistical Power Analysis for the Behavioral Sciences, 2nd ed. (Hillsdale, NJ: Erlbaum, 1988).

24. Ibid.

25. Ibid.

26. Ibid.

27. Graham Stone and Ellen Collins, "Library Usage and Demographic Characteristics of Undergraduate Students in a UK University," Performance Measurement and Metrics 14, 1 (2013): 25–35, doi:10.1108/14678041311316112.

28. Angie Thorpe, Ria Lukes, Diane J. Bever, and Yan He, "The Impact of the Academic Library on Student Success: Connecting the Dots," portal: Libraries and the Academy 16, 2 (2016): 373–92, doi:10.1353/pla.2016.0027.

29. Krista M. Soria, Shane Nackerud, and Kate Peterson, "Socioeconomic Indicators Associated with First-Year College Students' Use of Academic Libraries," Journal of Academic Librarianship 41, 5 (2015): 636–43, http://www.sciencedirect.com/science/article/pii/S0099133315001111.

30. Ellen Collins and Graham Stone, "Understanding Patterns of Library Use among Undergraduate Students from Different Disciplines," Evidence Based Library and Information Practice 9, 3 (2014): 51–67; Thorpe, Lukes, Bever, and He, "The Impact of the Academic Library on Student Success." [End Page 139]

31. Cox and Jantti, "Capturing Business Intelligence Required for Targeted Marketing, Demonstrating Value, and Driving Process Improvement"; Soria, Fransen, and Nackerud, "Beyond Books"; Graham Stone and Bryony Ramsden, "Library Impact Data Project: Looking for the Link between Library Usage and Student Attainment," College & Research Libraries 74, 6 (2013): 546–59, doi:10.5860/crl12-406.

32. Francesca Marineo and Heather Christensen, "Modeling ID Principles for Sustainable Instruction and Collaboration: Making the Library Integral to the LMS and Campus Processes," in Brandon West, Kimberley D. Hoffman, and Michelle Costello, eds., Creative Approaches to Instructional Design in Libraries: Moving from Theory to Practical Application (Chicago: ACRL, 2017).

33. Collins and Stone, "Understanding Patterns of Library Use among Undergraduate Students from Different Disciplines"; Jantti and Cox, "Measuring the Value of Library Resources and Student Academic Performance through Relational Datasets"; Soria, Fransen, and Nackerud, "Library Use and Undergraduate Student Outcomes"; Soria, Fransen, and Nackerud, "Beyond Books"; Stone, Ramsden, and Pattern, "Library Impact Data Project Toolkit"; Stone and Collins, "Library Usage and Demographic Characteristics of Undergraduate Students in a UK University." [End Page 140]

Additional Information

ISSN
1530-7131
Print ISSN
1531-2542
Pages
117-140
Launched on MUSE
2018-01-04
Open Access
No
Back To Top

This website uses cookies to ensure you get the best experience on our website. Without cookies your experience may not be seamless.