INTERDEPENDENCE OF THEORY, METHODOLOGY, AND EMPIRICAL DATA - CAUSAL MODELING AS AN APPROACH TO CONSTRUCT VALIDATION (FROM LONGITUDINAL RESEARCH ON DRUG USE, 1978, BY DENISE B KANDEL - SEE NCJ-50619)

• NCJ Number: 50631
• Author(s): P M BENTLER
• Date Published: 1978
• Length: 36 pages
• Grant Number(s): DA01070
• Sponsoring Agency: US Dept of Health, Education, and Welfare
• Corporate Author: Hemisphere Publishing Corp
• Publication Format: Document
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America

Annotation

A CAUSAL-MODELING APPROACH TO CONSTRUCT VALIDATION IN LONGITUDINAL STUDIES IS PROPOSED IN A DISCUSSION CONCERNED WITH THE INTERDEPENDENCE OF SUBSTANTIVE THEORY, THEORY OF METHODOLOGY AND STATISTICS, AND EMPIRICAL DATA.

Abstract

SUBSTANTIVE THEORY IS LITTLE MORE THAN SPECULATION WITHOUT SOUND DATA GATHERED IN THE CONTEXT OF AN APPROPRIATE METHODOLOGY TO EXPLOIT BOTH THEORY AND DATA. WITHOUT AN IMPORTANT PROBLEM TO TACKLE AND WITHOUT SUBSTANTIVE THEORY TO GUIDE ANALYSES OF VIABLE EMPIRICAL DATA, THEORETICAL DEVELOPMENT IN METHODOLOGY, MATHEMATICAL MODELING, OR STATISTICS CANNOT RISE ABOVE THE PURELY FORMAL. EVEN THE BEST EMPIRICAL DATA REQUIRE ATTENTION TO THEIR INHERENT METHODOLOGICAL AND THEORETICAL STRENGTHS AND DEFICIENCIES. THESE INTERRELATIONS MAY BE EXPLORED IN THE CONTEXT OF CONSTRUCT VALIDATION WHICH INVOLVES THE SCIENTIFIC CONFIRMATION OF TESTS AND MEASURES AS INDEXES OF POSTULATED ATTRIBUTES OR QUALITIES. THE METHODOLOGICAL TOOLS OF CAUSAL MODELING (ALSO CALLED STRUCTURAL EQUATION MODELING, CONFIRMATORY FACTOR ANALYSIS, AND ANALYSIS OF COVARIANCE STRUCTURES) MAY BE USED TO OPERATIONALIZE AND ASSESS CONSTRUCT VALIDATION. CAUSAL MODELING IS A METHOD OF TRANSLATING THEORETICAL STATEMENTS ABOUT THE INFLUENCES OF CERTAIN SETS OF VARIABLES ON OTHER SETS OF VARIABLES INTO A COMPLETE SYSTEMS FRAMEWORK THAT YIELDS TESTABLE STATISTICAL CONSEQUENCES. A CAUSAL-MODELING APPROACH TO CONSTRUCT VALIDATION IS PARTICULARLY USEFUL IN LONGITUDINAL RESEARCH WHICH TYPICALLY DEALS WIH MULTIPLE INDICATORS OF VARIOUS IMPORTANT CONSTRUCTS. THE INDICATORS ARE MEASURED TO BE UNDERSTOOD IN THE CONTEXT OF VARIOUS BACKGROUND OR CONTROL VARIABLES. A CAUSAL-MODELING APPROACH ALLOWS THE ENTIRE SYSTEM OF LONGITUDINAL DATA TO BE ANALYZED SIMULTANEOUSLY IN THE CONTEXT OF THEORY. A PROPOSAL FOR USING CAUSAL MODELING TO IMPLEMENT CONSTRUCT VALIDATION IS PRECEDED BY A REVIEW OF THE ROLES OF DESCRIPTION AND EXPLANATION IN SOCIAL SCIENCE. THE REVIEW INVOLVES A DISCUSSION OF THE METHODOLOGICAL APPROACHES--DESCRIPTIVE, EXPLORATORY AND MODEL BUILDING, AND HYPOTHESIS TESTING--IN SUBSTANTIVE RESEARCH, INCLUDING AN ASSESSMENT OF THE INTERACTION OF DATA AND THEORY IN EACH APPROACH AND A REVIEW OF THE CONCEPT OF CONSTRUCT VALIDITY. DEVELOPMENTS IN THE ANALYSIS OF MEAN AND COVARIANCE STRUCTURES THAT COMBINE THE BEST OF PSYCHOMETRIC, ECONOMETRIC, AND STATISTICAL THEORY ARE ALSO NOTED. THE DISCUSSION CLOSES WITH A CONSIDERATION OF THE GENERALITY AND SCOPE OF MATHEMATICAL MODELS AS THEY INTERRELATE WITH SOCIAL SCIENCE THEORY AND DATA. ANALYSES OF LONGITUDINAL DATA FROM STUDIES IN DRUG ABUSE AND OTHER FIELDS ARE CITED. A LIST OF REFERENCES IS INCLUDED. (AUTHOR ABSTRACT MODIFIED--LKM)