INVARIANCE OF MULTIVARIATE RESULTS - A MONTE-CARLO STUDY OF CANONICAL FUNCTION AND STRUCTURE COEFFICIENTS

Monte Carlo methods were employed to evaluate the degree to which canonical function and structure coefficients may be differentially sensitive to sampling error. Sampling error influences were investigated across variations in variable and sample sizes, across variations in average within-set correlation sizes, and in across-set population correlation sizes. Sixty-four different research situations were investigated, and for each situation 1,000 random samples were drawn. Results suggest that both sets of coefficients are roughly equally influenced by sampling error, except perhaps when some intradomain correlation coefficients are quite large. Thus, the case for emphasizing interpretation of structure coefficients must be made on a psychometric basis rather than on the grounds that structure coefficients are less sensitive to sampling error influences, notwithstanding the positions previously taken by some theorists. © 1991 Taylor & Francis Group, LLC.