The three methods have in common that the psychiatric disorder is measured with a set of symptom endorsements or questionnaire items. Taxometric procedures, model-based clustering and LVMM are different statistical approaches, and are based on assumptions that require different types of a priori knowledge about the collected data. As discussed in more detail later, taxometric procedures use covariances and therefore assume linear relationships between the items. Model-based clustering requires the user to choose a specific distribution for the items, and when fitting latent variable mixture models the user has to specify factor models as well as choose a distribution. As the performance of a method generally depends on whether its assumptions are adequate for the data, the choice of a method for a specific analysis should be guided not so much by what is common in a particular area of research but by whether the user is comfortable making the assumptions about the data that are required by the chosen method.
