THE STUDY OF SAMPLING ERRORS IN FACTOR ANALYSIS BY MEANS OF ARTIFICIAL EXPERIMENTS.

2 POINTS OF VIEW ABOUT FACTOR ANALYSIS ARE OFFERED: (1) THE STATISTICAL, IN WHICH SAMPLING FLUCTUATIONS ARE THE RESULT OF THE SAMPLING OF PERSONS, AND (2) THE PSYCHOMETRIC, IN WHICH INDETERMINACIES RESULTING FROM FLUCTUATIONS IN WHAT IS MEASURED ARE EMPHASIZED. SEVERAL STUDIES SUGGEST THAT THE STANDARD ERROR OF A FACTOR LOADING DUE TO THE SAMPLING OF PERSONS IS ABOUT N-$M$UQ. THIS IS APPROXIMATELY TRUE FOR ANALYTICALLY ROTATED FACTORS. IT IS MORE NEARLY TRUE FOR LEAST-SQUARES ROTATIONS TO THE POPULATION, BUT THIS GENERALIZATION IS LIMITED BY THE LACK OF CONTROL OF THE ANGLES OF SEPARATION AMONG THE ROTATED FACTORS IN THE SAMPLE. EVIDENCE IS GIVEN CONCERNING THE INFLUENCE OF TYPE OF FACTORING, SIZE OF LOADING, AND METHOD OF FACTORING. THE DETERMINATION OF THE NUMBER OF FACTORS IS DISCUSSED. EVIDENCE FROM STUDIES OF PSYCHOMETRIC ERROR IS ALSO PRESENTED. (39 REF.) (PsycINFO Database Record (c) 2010 APA, all rights reserved)