Robust Clustering with Subpopulation-specific Deviations
share
The National Birth Defects Prevention Study (NBDPS) was a case-control study of birth defects conducted across 10 U.S. states. Researchers are interested in characterizing the etiologic role of maternal diet, using data tools such as the food frequency questionnaire. Maternal diet and behaviors have been shown to influence the development of congenital malformations. In a large, heterogeneous population, traditional clustering methods, such as latent class analysis, used to estimate dietary patterns can produce a large number of clusters due to a variety of factors, including study size and regional diversity. These factors result in a loss of interpretability of patterns that may differ due to minor consumption pattern changes. Based on adaptation of the local partition process, we propose a new method, Robust Profile Clustering, to handle these data complexities. Here, participants may be clustered at two levels: (1) globally, where women are assigned to an overall population-level cluster via an overfitted mixture model, and (2) locally, where regional variations in diet are accommodated via a beta-Bernoulli process dependent on subpopulation differences. We use our method to analyze the NBDPS data, deriving pre-pregnancy dietary patterns women in the NBDPS while accounting for regional variability.
READ FULL TEXT
