Calculated Clustering-An Application to Childhood Growth Trajectories

Tuesday, December 4, 2012 at 1:00pm to 1:45pm

Bronfman Science Center, 106 18 Hoxsey St, Williamstown, MA 01267, USA

Calculated Clustering-An Application to Childhood Growth TrajectoriesBrianna Heggeseth, University of California, Berkeley
Mathematics and Statistics Department Colloquium

Abstract: Through the integration of technology into practically every aspect of our daily lives, it is becoming increasingly possible to collect massive amounts of data on individuals over time. My work revolves around finding meaning and structure in these large longitudinal data sets. One way to discover structure is through calculated clustering of the data. Implicit in any grouping of objects is a definition of similarity/dissimilarity between the objects, which is highly dependent on the scientific research question. I will introduce two standard clustering approaches, K-means and Finite Mixture Models, used for longitudinal data and illustrate how these algorithms indirectly define similarity. Then I will propose a new clustering method that focuses on a different definition of similarity. Throughout the talk, these methods will be discussed with reference to childhood growth patterns of individuals in the Center for the Health Assessment of Mothers and Children of Salinas (CHAMACOS) study.

Event Type



Mathematics & Statistics

Recent Activity

People Going

Getting Here