Classification accuracy ranged from 60% to 76%. While these classification rates are statistically significant, both exposed and non-exposed subjects were misclassified. To determine if there were any systematic differences accounting for the misclassification, we compared subjects that were misclassified to those that were correctly classified for each analysis described above. Two continuous variables, age and FSIQ, were tested using ANOVA. The following categorical variables were tested by Fisher’s exact test or chi-square: sex, handedness, country (U.S. vs. South Africa), race, ethnicity, growth deficiency (height or weight ≤ 10%), height ≤ 10%, weight (≤10%), microcephaly (OFC ≤ 10%), structural abnormality (two or more of the following facial features: short palpebral fissure length, smooth philtrum, thin vermillion border), short palpebral fissures, smooth philtrum, thin vermillion border, ADHD diagnosis. Presence or absence of physical features were analyzed individually as well as grouped into the structural abnormality and growth deficiency variables. See Table 2 for descriptive data and definitions of the variables included.
