| Sumario: | The article presents a reply in response to a comment on an article that discusses about group variables. According to the author most sensitive analysts have found it difficult to cope with the multiplicity of group properties. The author thanks professor Alan B. Wilson, of the University of California, Berkeley, for drawing attention to a problem in the approach when the number of groups is smaller than the number of variables. Although factor analysis (the determination of the relations between the items and the factors) can proceed in a straightforward manner when this is the case, the computation of factor scores for each group cannot. The exact computation of factor scores involves the inversion of the matrix of correlations between the variables; this cannot be done when the number of variables is smaller than the number of groups, because such a matrix is necessarily "singular" some correlations can be expressed as linear combinations of the other correlations. Now that computer programs for obtaining factor scores are available, it will be essential to have fewer variables than groups.
|
|---|