Research by IR Office Staff
The Institutional Research Office staff periodically conducts research for presentations at regional and national institutional research conferences. This page presents some of these results and presentations.
Nonresponse rates on student surveys
College campuses are beehives of opinion surveys that aim to inform institutional decision making on curriculum changes, student experiences, assessment of programs and strategic initiatives, or even students’ food choices or naming the mascot. Even though technology such as Google Forms has made it easier to conduct surveys, we suspect there is not enough attention to who is responding to our surveys and who is not. In other words, response rate issues may lead to non-response bias, meaning that the opinions of those who respond differ from those who do not. We wanted to learn whether there are differences in response rates between certain subgroups of students who respond to a survey or who do not, such as males or females, minority or non-minority. We conducted a national study on assessing and addressing student non-response in national liberal arts colleges and universities. In this study, we discuss methodological improvements such as post-stratification weighting and best practices such as community engagement to improve the response rate in student surveys. We showcase how such enhancements improved the response rate of a senior survey we recently administered. Our study includes novel approaches and ideas of institutional research professionals, and it is likely to advance one of the profession’s core functions—conducting student surveys. We presented this research at the Connecticut Association of Institutional Research (ConnAIR) meeting in May 2023 and at the North East Association of Institutional Research annual conference in Novemner 2023.
View the presentation (PDF): Assessing and Addressing Survey Nonresponse - NEAIR 2023
Development and Use of a Peer Group in Institutional Research
Colleges and universities use comparison groups or “peer groups” for a variety of analytical purposes. A peer group consists of schools that are "like” one’s own institution in specifiable ways and can be used to clarify the college’s market position, to allow comparison of activities and outcomes, and to benchmark management decisions regarding faculty/staff salaries, divisional budgets, staffing levels, and the like.
Selecting a peer group is a combination of creativity, science, and politics. Several statistical techniques, such as factor analysis, cluster analysis, regression, and discriminant analysis, can be deployed to generate a peer list. Before deciding what method to use, a researcher must carefully select the population of schools to draw the list and the variables of interest. After that, the dataset should be cleaned to eliminate unusual cases, and the variables must be vetted for relevance to the context in which they are being used.
Cluster analysis, one of the widely used statistical techniques for selecting a peer group, takes center stage in our study. We delve into the research behind it, its functioning, and the most effective ways to leverage it. Finally, we demonstrate the practical application of cluster analysis by using it to select a peer group for Connecticut College. This study was presented at the annual Connecticut Association of Institutional Research (ConnAIR) conference, held at the University of Connecticut in May 2024.
View the presentation (PDF): Development and Use of a Peer Group in IR