American Statistical Association
Measures of biological phenomena, in general, and of psychiatric conditions, in particular, often exhibit symmetric shapes resembling a normal distribution. However, it is well known that members of homogeneous populations with symmetric unimodal distributions can exhibit very distinct characteristics. In medical research, an important goal is to identify groups of subjects characterized by a particular trait or quality and to distinguish them from other subjects in a clinically relevant way.
Statistical approaches predominantly used for that purpose have been based on the assumption of underlying categories, whether observed or latent, such as latent class or mixture and growth mixture models. In this talk, I will present an approach to characterizing heterogeneity, which is based on principal points methodology, that does not presuppose the existence of distinct categories. The approach can be used with any multivariate distribution, but in this talk, I will present the case of functional (longitudinal) data modeled with mixed effects models.
As an illustration, the following clinical question will be addressed: After a depressed subject is treated with an antidepressant and responds to the acute treatment, can we tell whether s/he still needs the drug to maintain response? Data from two discontinuation depression studies will be used. The studies were designed and carried out by investigators from DES at NYSPI and Columbia U.
|Date:||Tuesday, November 17, 2009|
|Time:||3:00 - 4:00 P.M.|
New York State Psychiatric Institute
1051 Riverside Drive
6th Floor Multipurpose Room (6602)
New York, New York