Home > Events > 2012 Seminars & Colloquia > AMY HERRING, The University of North Carolina at Chapel Hill

AMY HERRING, The University of North Carolina at Chapel Hill

Main Content

Bayesian latent class models for complex data
When
31 January 2012 from 4:00 PM to 5:00 PM
Where
201 Thomas Bldg.
Add event to calendar
vCal
iCal

For clinical guidelines, simplicity is important.  For this reason, epidemiologists and clinicians often categorize continuous predictors before analysis.  For a time-varying or high-dimensional predictor, predefining categories is not straightforward.  Latent class analysis provides an attractive approach for categorization through clustering.  We develop a semiparametric Bayes approach that avoids assuming a prespecified number of clusters and allows the response to vary nonparametrically over predictor clusters.  The methodology is motivated by interest in relating trajectories in weight gain during pregnancy to the distribution of birth weight, adjusted for gestational age at delivery.  The proposed approach allows the tails of the birth weight density to vary flexibly over weight gain clusters.  Recent extensions and future directions are described.

Filed under: ,