Table 1 shows that some of the categories for the Teacher's Advice and Father's Profession variables are relatively empty. So, for the Teacher's Advice variable we put "No further education" and "Extended primary education" into a single category, and similarly for "Manual" and "Agricultural", and "Secondary school" and "Pre-University", resulting in four categories. For Father's Profession we put together "Unskilled Labour", "Schooled Labour" and "Lower white collar", again ending up with four categories.
A two-dimensional HOMALS-solution for all the variables shows a fit of .80, which is quite high. This solution also reveals many interesting features of the data.
In Graph 1, the plots give the original vs the optimally transformed category quantifications in both dimensions for the IQ, Teacher's Advice and Father's Profession variables. The transformations are monotone on the first dimension for all the variables. The violations of ordinality on the second dimension mainly occur in the higher categories of all three variables.
The category quantifications of the Gender, IQ, Teacher's Advice and Father's Profession variables are presented in Graph 2 . It can be seen that the first dimension discriminates really well Father's Profession, while both dimensions discriminate equally well variables such as IQ and Teacher's Advice. In order to examine the stability of the solution, bias corrected confidence intervals for the category quantifications of a one-dimensional HOMALS-solution were computed and presented in Graph 3.
The discrimination measures are shown in Graph 4. This graph shows that the first dimension is related to the Father's Profession variable. Both dimensions discriminate equally well the variables IQ, Teacher's Advice and School. The variables with the largest discriminating power are Teacher's Advice and IQ. On the other hand, Gender discriminates rather poorly.
The object scores are given in Graph 5. The most important point is that the object scores of Boys and Girls form two somewhat separate clusters. This finding suggests that a separate analysis for Boys and Girls should prove useful. Moreover, there is some indication for the presence of a horseshoe in this graph, indicating that we are dealing with a somewhat dominant first dimension on which the items are well ordered. This suggests that the use of the PRINCALS technique will overcome this problem.