Many constructs in abnormal psychology and psychiatry, such as depression or neuroticism, are not measured directly but by means of questionnaires or symptom collections. The construct itself is unobserved, or latent (see the online Supplementary material for more detail on representing disorders as latent variables). The individuals’ scores on the latent variable are thought to predict the responses on the symptoms (see Fig. 1a). The central question is whether the latent variable is categorical (i.e. a latent class variable), or continuous (i.e. a latent factor). A latent class variable categorizes a sample into groups (e.g. affected/unaffected or subtypes), and group membership predicts how likely it is that a person will endorse a symptom. By contrast, a latent factor aligns individuals in a sample on a continuum, and the higher a person scores on the factor, the more likely they will endorse a symptom. If the construct is categorical and can be represented by a latent class variable, then the observed items covary because of mean differences between the groups or classes.
