Parental data used to predict growth of craniofacial form

The sample for this study consisted of 250 families. From 850 lateral and posteroanterior cephalograms, 81 variables from the X, Y coordinates of 67 landmarks were calculated and matched with the normal distribution. Principal component analysis was used to summarize these 81 variables in proper factor scores. The craniofacial patterns of 500 adults were classified by cluster analysis on the basis of those factor scores. This study introduced the concept of "similar parent" and "dissimilar parent" instead of the father and mother equally. Finally, the following model for predicting the individual growth of craniofacial characters was obtained. The predicted value Y(t) [formula: see text] is the logistic curve used in growth studies, C1X(s) is for the similar parent of a child, and C2X(d) is for the dissimilar parent. Multiple regression functions were calculated for both sexes in each of four craniofacial patterns. The errors of prediction at the average age of 18 years from the data at the average age of 11.6 years were 1.12, 2.88, 2.87, 3.14, and 1.93 mm for the respective distances S-N, N-Me, S-Me, Gn-Cd, and G-G'. These errors in our growth prediction are much smaller than in ordinary normal facial diagrams and may be considered negligible for orthodontic clinical application.