|
Abstract:
|
In most of clinical trials, patients are monitored over time producing a longitudinal response profile for each patient. We are interested in assessing whether there are group of patients that exhibit some common profiles. For example, patients could be clustered as responders and non-responders. Once we have identified clusters and their response profiles, we intend to use baseline characteristics, medical histories or genomic information to help us find prognostic factors for a given cluster of patients. This latter effort is a secondary step in the process; it is first important to identify whether, how many and what kind of response profiles may exist in the data. In this paper, we propose a new Bayesian growth mixture model to identify latent clusters. The advantages of this Bayesian approach for tailoring are demonstrated through examples.
|
