Briana Joy K. Stephenson, PhD, Harvard T.H. Chan School of Public Health
Title:听Bayesian nonparametric solutions to analyze nutritional Survey Data.
础产蝉迟谤补肠迟:听奥别产蝉颈迟别:
Dietary intake is a major modifiable risk factor for cardiovascular disease that has a disproportionate impact on low-income and racial/ethnic minorities. Population-based studies provide researchers with a snapshot of dietary habits of a target population through the collection of dietary intake assessments (food frequency questionnaire, 24-hour dietary recalls) from a large sample of participants. Through the implementation of complex survey designs and recruitment strategies, researchers can obtain a diverse random sample of a target population to better understand the larger target population of interest. Characterization of dietary intake from these assessment tools can often be quite complex due to the high-dimensional structure of the data. When analyzing national multistage survey data with unequal probabilities of selection and response inherent in the design, an additional layer of complexity is presented.
Bayesian nonparametrics offers a more efficient solution that can accommodate (1) complex high dimensionality of dietary intake data, (2) volume of a large population size, (3) preserve model stability in the presence of sparsely consumed foods, and (4) integrate prior information with observed data. Using diet consumption data collected in large population-based survey cohorts, this talk will discuss flexible solutions Bayesian nonparametric model-based clustering techniques can provide to manage the complexities of dietary heterogeneity present in populations typically understudied and marginalized (e.g. low-income or racial/ethnic minority) in the United States.
Please visit our website for the Zoom Link: /epi-biostat-occh/seminars-events/seminars/biostati...
听
听