CLASSIFICATION AND REGRESSION TREE ANALYSIS OF A NEUROLOGICALLY IMPAIRED AND NORMAL SAMPLE USING SENSORY-MOTOR TASKS

Abstract

The ability to differentiate between neurologically impaired and normal individuals is an important component in a valid neuropsychological battery. However, limited research exists regarding the ability of sensory-motor batteries to differentiate between the two groups. This study used Classification and Regression Tree Analysis (CART) to identify which measures of sensory-motor functioning from the Dean-Woodcock Sensory Motor Battery (DWSMB) would best differentiate between neurologically impaired and normal individuals, as well as identify which subtests would provide the best pathognomic power. The results revealed that a number of clinically useful nodes emerged that enabled the differentiation between groups with a small number of tasks. The primary separation variable was the Gait and Station subtest, a measure of subcortical motor functioning. Auditory Acuity and Clock Construction also provide important pathognomic information. A cross validation was conducted to determine the integrity of the generated decision tree, and results revealed that the generated model correctly predicted 84.5% of the normal group and 71.4% of the neurologically impaired sample. The results from the present analysis provides further evidence for the construct validity of the DWSMB.