Discriminant analysis in the classification of anxiety disorders

The urgency of preventing serious mental disorders (MDs) has intensified in recent decades demanding innovative approaches for early diagnosis. This paper’s main objective is to revisit statistical discriminant methods emphasizing their crucial and practical role to classify patients into different MDs categories. From three groups (nervous, psychotic, and healthy) of fifty individuals, each evaluated by thirty variables, an exploratory discriminant analysis was performed in order to obtain the linear combination of the variables who maximize the separation between these groups. From the statistical analysis of the two first discriminant functions, it was identified a subset of fifteen variables which discriminant power revealed a misclassification rate of 10% in the training test and 14.6% in the testing test. Finally, this model was compared to a discriminant stepwise method which identified eighteen discriminant variables.