Answers are grounded in this paper and cite their sources.
A common problem in multivariate applications is the comparison of pattern matrices obtained from two independent studies. We compared the performance of four pattern matching indices (the coefficient of congruence [c], the s-statistic [s], Pearson's r [r], and kappa [k]) under a variety of experimental conditions. We constructed population pattern matrices by varying (a) saturation or the size of the loadings, (b) sample size, (c) the number of observed variables, and (d) the number of derived variables. Sample patterns were computer generated and matched, employing each index, to their population pattern. With the exception of r, little difference in matching performance between indices was observed. In general, an increase in either saturation or sample size resulted in more accurate index values.
No takes yet. Share an insight, caveat, or question.
Guadagnoli et al. (1991) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: