PDF pending restoration
ERIC Number: ED227181
Record Type: RIE
Publication Date: 1982-Apr
Reference Count: 0
Images or Aberrations? Human Judgment and Insight as Reflected In Current Regression Analyses.
In contrast to linear models of human judgment developed for predictive purposes which are characteristically insensitive to the exact values of the weights utilized in them, the Linear Multiple Regression Models used for policy capturing are assumed to reflect significant aspects of the subjects' judgmental policies. This latter kind of modelling is therefore justified only to the extent that its underlying assumption is found to be true. Its validity depends, however, on both that of the models' beta weights, and of R2 as a twin measure of the subjects' cognitive control/consistency and of one's success in capturing the judges' policy. Although the problematic nature of beta weights is well known, the present study shows that R2 is no less a problematic measure. It is shown that the way in which self-insight is typically elicited may induce a demand-response effect; furthermore, traditional data analysis and comparison methods appear to be inconsistent. In the study, eight subjects were presented with 72 profiles of perspective undergraduate students and were asked to judge future academic performance, based on information regarding each student's sex, age, ethnic oigin, and socioeconomic and educational background. The results indicate that current policy capturing research and findings cannot be accepted at face value. (Author/GC)
Descriptors: Evaluative Thinking, Higher Education, Judgment Analysis Technique, Multiple Regression Analysis, Performance Factors, Policy Formation, Predictor Variables, Statistical Bias, Student Characteristics, Validity
Michael Inbar, Department of Sociology, Hebrew University, Mount Scopus, Jerusalem 91905, Israel (write for price).
Publication Type: Reports - Research
Education Level: N/A
Sponsor: National Council of Jewish Women, New York, NY. Research Inst. for Innovation in Education.
Authoring Institution: N/A
Identifiers: Beta Weights; R2 Values