An Application of Discriminant Functions to the Problem of Predicting Brain Damage Using Behavioral Variables

Publisher Website
Google Scholar
Abstract
Discriminant functions were applied to 24 behavioral indices from each of 140 individuals tested in the Neuropsychology Laboratory, Indiana University Medical Center. The scores came from 11 Wechsler subtests, 11 Halstead subtests, and 2 Trail Making tests. Comparisons were made between 61 non-brain-damaged control Ss and 79 Ss showing any of several kinds of cerebral damage on the basis of criteria resulting from independent neurological examination. These 79 were subdivided and compared with the controls as follows: (a) 25 Ss with damage of the left cerebral hemisphere, (b) 31 Ss with damage of the right hemisphere, (c) 23 Ss with diffuse, bilateral involvement. Also the 25 left-damaged and the 31 right-damaged Ss were compared. The discriminant function in each comparison produced a single weighted score per S, an optimum, least-squares type of separation between the two groups. The resulting distributions of summed, weighted scores in each comparison were inspected for the point of minimum overlap. An individual's weighted score, falling above or below this point, categorized him as belonging to one group or the other. These assignments, when compared with the criterion assignments of Ss to groups, were expressed as percentages of correct predictions: controls versus all categories of cerebral damage, 90.7%; controls versus left damage, 93.0%; controls versus right damage, 92.4%; controls versus diffuse damage, 98.8%, and right versus left damage, 92.9%. Several other indices were examined for percentages of correct prediction, but the discriminant function was superior in all comparisons. A cross-validation study is now in press (6).

This publication has 3 references indexed in Scilit: