Sparse Functional Data with an Irregular Periodic Component: An Application to Psychiatric Data

RAND Statistics Seminar Series

Sparse Functional Data with an Irregular Periodic Component: An Application to Psychiatric Data

Dr. Catherine A. Sugar—UCLA

Thursday, November 10th, 2011
10:30 a.m. – 12:00 p.m. PT
Conference Room 4144
RAND Corporation, Santa Monica, CA

Other Locations/Times:
Washington, D.C., Conf. Rm. 4132 1:30 p.m. ET
Pittsburgh, PA, Conf. Rm. 5202 1:30 p.m. ET

Please contact Fabiola Lopez if you would like to attend this seminar.


Many psychiatric illnesses have a periodic component, with patients experiencing illness (episodes) of varying length and severity, interspersed with periods of relatively good health. Even when subjects are followed longitudinally over an extended time frame the resulting data are often difficult to analyze, both because of irregularity in the amplitude, period length and sequencing of events, and because observations tend to be sparse relative to the natural illness cycle. In this talk, we present several approaches to dealing with such data. The first is a state-based approach in which the raw functional data are used to identify when episodes are occurring and a combination of Markov chain and bootstrapping techniques are then used to examine subjects’ transitions between states over time. The second, more model-based approach focuses on transforming the raw data into functions for the amplitude, rate and length of episodes which can then be analyzed using standard functional data analysis techniques. In this framework, sparsity can be dealt with by borrowing strength across subjects using random effects models, by jointly examining trajectories of multiple outcomes per subject, and by using pre-treatment or other supplementary data to better assess within subject periodicity. The approaches will be illustrated with an application to bipolar disorder where the goal is to look at differential treatment effects on length, frequency and severity of manic and depressive episodes.

Speaker Bio

Dr. Sugar is an Associate Professor in the Departments of Biostatistics and Psychiatry & Biobehavioral Sciences at the University of California, Los Angeles. She also directs SIStat, the biostatistics core of the Semel Institute for Neuroscience and Human Behavior. Dr. Sugar received her Ph.D. in Statistics from Stanford University in 1998, where she worked on developing cluster analytic methodologies for defining health state models with applications. In addition to clustering, her methodological interests include functional data analysis, classification and high dimensional multivariate data analysis. Dr. Sugar has worked on studies in a wide range of areas including health services research, dentistry and HIV/AIDS, but her primary application area is mental health. She is currently the director of the statistics and data management cores for multiple psychiatry and neuroscience centers at UCLA involving schizophrenia, bipolar disorder, ADHD, and autism.

Attending a Seminar

RAND visitors are welcome to attend and must RSVP at least one day prior to the seminar. To ensure your attendance please contact Fabiola Lopez at with your name, company (or university) affiliation, and national citizenship (for security purposes).

For parking and directions to RAND's Santa Monica office, please see:

For parking and directions to RAND's Pittsburgh office, please see:

For parking and directions to RAND's Washington office, please see:

For further information and to be added to the mailing list contact Fabiola Lopez at