The predictive validity of multidimensional psychological spaces
Cornelius, Edwin Thomas
Cornelius, Edwin Thomas
item.page.creator
Citations
Altmetric:
Soloist
Composer
Publisher
Date
1973
Additional date(s)
Abstract
The purpose of the present research was to test a paradigm for computing the predictive significance of multidimensional psychological spaces within the context of college student preferences for the presidential candidates in the 1972 Democratic primaries. An experiment was conducted in which subjects scaled a set of personality adjective traits and a set of political issues. A spatial structure was derived for each space using Carroll and Chang's INDSCAL computer program. Voters and presidential candidates were embedded in the original spaces using Model II of Carroll's Linear-Quadratic regression procedure. Distances between voters and candidates in the joint spaces were computed using six different distance models. The results were then compared to independently collected voting preferences for the political candidates. Although some of the spatial models did better than chance in predicting the voting preferences of the subjects, the results were generally unimpressive. Several explanations for the negative findings were offered, and a discussion of the implications of the present findings for future work in MDS validation was given. In addition, issues of secondary interest to the research were discussed, including the problem of interpreting negative dimensional weights that were obtained from Carroll's quadratic-regression procedure.
Contents
Subject
Subject(s)
Psychometrics
Research Projects
Organizational Units
Journal Issue
Genre
Dissertation
Description
Format
vi, 132 leaves, bound : illustrations
Department
Psychology