ABSTRACT
Refinement was usually accomplished by small graphic rotations of the axes taken two at a time. The method is a slight modification of L. L. Thurstone's analytic single-hyperplane procedure. No initial rotation, whether it is based on an entirely numerical procedure such as the Optres transformation or on the examination of a plot as in the extended-vector procedures, leads to the best attainable simple structure. When the objective of a factor analysis is merely to interpret the rotated factors and to judge the excellence of the simple structure by noting the overdetermination of each factor, the analysis is ordinarily concluded with the Promax or Optres transformation. But when the investigator wants really excellent estimates of the primary factors, and especially when he intends to go on to second-order analysis, hierarchical analysis, and/or the computation of factor scores, further refinement of the simple structure on the reference vectors is advisable.
