Date of Award
Doctor of Philosophy
Ed Psychology and Research
How college affects students is a central phenomenon of interest in higher education research. However, a major problem in assessing the influence of college on students is the methodological dilemmas due the multilevel nature of the majority of data used in such studies. Historically, higher education researchers have utilized the traditional linear model, ordinary least squares (OLS) regression, to aid in their investigation of the influence of college on students. This traditional approach ignores the multilevel nature of the data which can cause a multitude of conceptual and statistical problems. Therefore, a statistical technique, such as hierarchical linear modeling (HLM), that takes into account the multilevel nature of the organization of higher education is need. The purpose of this study is to determine whether conclusions regarding the influences on college seniors’ critical thinking ability would differ depending upon the type of analysis, OLS regression or the more appropriate HLM analysis. In this study, the influences on seniors’ critical thinking ability is examined three ways— (1) an OLS regression with the student as the unit of analysis, (2) an OLS regression with the institution as the unit of analysis, and (3) a three-level HLM with student attributes modeled at Level 1, characteristics of the major modeled at Level 2, and characteristics of the institution modeled at Level 3— in order to illustrate the differing conclusions one may come to depending upon the type of analysis chosen. Overall, evidence from this sample suggest that one would come to substantively different conclusions regarding the influences on students’ perceived critical thinking ability depending upon the type of analysis chosen, especially in regards to the effects of the institutional characteristics. Specially, the results from the institution-level OLS regression cannot be considered reliable. Findings from the institution-level OLS regression model differed substantially from the results of the other two analyses. The results from the student-level OLS regression analysis can only be partially trusted. The student-level OLS regression produced results comparable to the HLM estimates for the lower-level variables but substantively different results for the institutional characteristics. Thus, when institutional characteristics are of prime importance, one should perform an HLM analysis in order to be confident in the results obtained for the institutional effects.
dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Rocconi, Louis M., "Analyzing Multilevel Data: An Empirical Comparison of Parameter Estimates of Hierarchical Linear Modeling and Ordinary Least Squares Regression" (2010). Electronic Theses and Dissertations. 88.