ABSTRACT
In a common-factor analysis we have n measured variables, from which we derive m + n factors: m common factors and n unique factors. It is evident then, because we have more unknowns (factors) than measured variables, that scores on the m common factors can only be estimated. Thus factor scores are best determined when the variables are highly reliable and when each common factor has a number of high loadings. The method of estimating factor scores described in termed the complete regression method. If there is a substantial cluster of variables on each primary axis, with or without one or more clusters elsewhere, a rough approximation to factor scores may be obtained by averaging the standard scores on the variables of each such cluster. Arguments in favor of such methods seem to be based primarily on the proposition that estimates of scores on orthogonal factors should be orthogonal.
