THE USE OF THE COMPUTER TO GENERATE STATISTICAL TABLES FOR THE STUDY OF PERSONALITY TRAITS, A MONTE CARLO AND A LOGICAL ANALYSIS OF MULTITRAIT-MULTIMETHOD STATISTICS.

THIS WAS A TWO PART INVESTIGATION. THE FIRST PART WAS A MONTE CARLO (STATISTICAL) ANALYSIS, AND THE SECOND WAS A LOGICAL ANALYSIS OF MULTITRAIT-MULTIMETHOD VALIDITY. PART ONE SUCCESSFULLY GENERATED, FOR SMALL SAMPLE SIZES, EMPIRICAL DISTRIBUTIONS OF STANLEY'S F STATISTIC FOR TESTING TRAIT VALIDITY IN MULTITRAIT-MULTIMETHOD MATRICES. ZYZANSKI'S CORRECTION OF STANLEY'S STATISTIC WAS FOUND INAPPROPRIATE FOR SMALL SAMPLE SIZES. PART TWO EMPLOYED THE METHOD OF LOGICAL ANALYSIS TO DETERMINE THE SOUNDNESS OF THE FOUR CRITERIA PROPOSED BY CAMPBELL AND FISKE FOR DETERMINING TRAIT VALIDITY BY MULTITRAIT-MULTIMETHOD MATRICES. IT WAS CONCLUDED THAT ONLY CRITERION ONE, CONVERGENT VALIDITY, COULD BE CONSIDERED A "THEOREM" OF TESTING THEORY. THIS ANALYSIS QUESTIONED WHETHER SPECIFIC TESTS CAN BE VALIDATED OR INVALIDATED WHEN THE CRITERIA OFFERED TO DO THIS ARE NOT THEMSELVES "VALID" OR LOGICALLY NECESSARY. (PS)