Comparison of Conjoint Analysis, Multiple Regression Models with Person Vectors and Profile Analysis to Assess Important Factors Used to Select Colleges

Abstract

The purpose of this study was to investigate the relative effectiveness of the traditional conjoint analysis approach to the multiple regression approach that includes person vectors profiles analysis. It was expected that the more sophisticated models would increase the effectiveness in terms of its shrinkage estimates and the accuracy of its predictability of two holdout groups. The data source consisted of a sample of 100 students who rated eight colleges on five attributes--quality of education, financial aid, quality of dorm life, student/faculty relations, and social aid.