The Methodology of Research on College Impact, Part Two

This first in a two-part series of papers is concerned with the following aspects of the methodology of research on college impact: different types of inferential errors, single-institution versus multi-institutional studies, longitudinal versus cross-sectional data, alternative statistical designs, the effects of measurement errors, and methods of detecting student-environment interaction effects. The most definitive information about college impact is obtained from multiinstitutional longitudinal studies in which data are obtained on student inputs, student outputs, and environmental characteristics. Although there are several multivariate designs available for dealing with such data, stepwise linear multiple regression analysis (with appropriate adjustments for measurement errors) is perhaps the most versatile method, particularly if the regression is carried out in separate stages dictated by the logic of the college impact process.