Grade of Membership Analysis of Depression-Related Psychiatric Disorders
- 1 August 1989
- journal article
- Published by SAGE Publications in Sociological Methods & Research
- Vol. 18 (1) , 126-163
- https://doi.org/10.1177/0049124189018001005
Abstract
Information on 7,184 respondents of the Duke University and Johns Hopkins University Epidemiologic Catchment Areas (ECA) studies were analyzed using a fuzzy set classification procedure called Grade of Membership (GoM) analysis. A formal description of GoM analysis and its relationship to other multivariate procedures, some of which are described in other articles in this issue, is provided. Symptoms were selected for analysis that represented seven clinically defined psychiatric disorders: depression, generalized anxiety, panic disorder, simple phobia, social phobia, agoraphobia, and somatization. In the sample there were persons who expressed symptoms for these diseases as well as a large group of persons free of any symptoms. In the GoM analyses, we used both the 41 symptom variables initially provided to define disorder groups and a derivative set of 33 symptom variables with certain of the original symptoms grouped to increase the stability of symptom sets. In the analysis it was found that six pure types best explained the variation of both the sets of 41 and 33 symptoms. The structure of the 33 variable set provided a better identification of the underlying disease groups. The six pure types that emerged from the 33 variable analysis were identified as depression, generalized anxiety, phobias, somatization, and two groups that were initially classified as asymptomatic persons. One of these groups was clearly asymptomatic and the other possessed only mild symptoms not strongly indicative of any of the disorders. This group may represent a neurotic subpopulation. Perhaps a surprising result was the failure for panic to emerge as a separate disorder but instead to be associated with generalized anxiety. The failure to identify the phobias as separate disorders may be attributable to the small number of variables used to define them. In a previous analysis using a larger set of variables, simple and social phobias clearly separated.Keywords
This publication has 8 references indexed in Scilit:
- The Identifiability of the Proportional Hazard ModelThe Review of Economic Studies, 1984
- Simultaneous Factor Analysis of Dichotomous Variables in Several GroupsPsychometrika, 1981
- Do Psychiatric Patients Fit Their Diagnoses? Patterns of Symptomatology as Described with the BiplotJournal of Nervous & Mental Disease, 1979
- Mathematical typology: A grade of membership technique for obtaining disease definitionComputers and Biomedical Research, 1978
- A new look at the statistical model identificationIEEE Transactions on Automatic Control, 1974
- Clinical data representation in multidimensional spaceComputers and Biomedical Research, 1970
- Consistency of the Maximum Likelihood Estimator in the Presence of Infinitely Many Incidental ParametersThe Annals of Mathematical Statistics, 1956
- Multiple factor analysis.Psychological Review, 1931