Some statistical issues in psychiatric epidemiology research.

Like many other fields, psychiatric researchers rely heavily on the utilization of conventional epidemiologic designs and principles to pursue their scientific objectives. However, some of the issues, which to a large extent are specific to psychiatric disorders, have demanded that new designs and hence new statistical methods be developed to meet the challenges. These issues include, among others, uncertainty of diagnoses due to the lack of well-established biologic markers, high prevalence of comorbidity and lastly, etiologic heterogeneity. In this note, we will elaborate on some problematic aspects of psychiatric epidemiology that pertain to the above three issues. We will then briefly discuss the shortcomings of current statistical approaches and provide suggestions for possible alternatives, some of which may require further development.