engleski

The matching of two sets of factors by decomposition of variance approach

The usual approach, proposed by several authors in matching of two factor solutions obtained on two different samples measured by the same set of variables is the determination of the matrix where and are either matrices of correlations between variables and orthogonal or oblique factors or parallel projections on oblique factors, as reported by Fruchter (1966). When the comparison of oblique factor solutions takes place, the difference in values obtained by two definitions of A matrices can be substantial, so it is recommended to calculate the congruences of both pairs, the structure and pattern matrices. In this paper, a new approach is proposed where a single sintetic measure represents the matching of both pairs of A matrices. The matching measure is defined as cosine of the angle between vectors of proportions of variance explained by the factors.

factor solution; matrix; congruence

nije evidentirano