Delaware Chapter, ASA
Dinner Meeting: September 2011
Meeting location: Townsend Hall on the University of Delaware campus. Townsend Hall
Reneé H. Moore, PhD |
Department of Biostatistics and Epidemiology
University of Pennsylvania
Perlman School of Medicine
Constrained Bayes Estimates of Random Effects
when Data are Subject to a Detection Limit
slides and photos.
In many environmental and HIV epidemiologic studies, the prediction of random effects corresponding to subject-specific characteristics (e.g., mean exposure or HIV viral load levels) is of much interest. A common feature of data from such studies is that some values may fall below a limit of detection associated with the measuring instrument. A widely accepted predictor of subject-specific random effects is the empirical Bayes (or posterior mean) estimate, which has the well-known potential disadvantage of overshrinking toward the population mean. Alternative "constrained Bayes" predictors maintain favorable properties while reducing this shrinkage. Although the Bayes estimate has been adapted to predict random effects when data contain non-detectable values, the constrained Bayes methodology has not been demonstrated in this setting. In this talk, we combine general methods found in the literature for computing constrained Bayes estimates with methods for computing Bayes estimates in the presence of non-detects. The resulting constrained Bayes predictors are compared with the Bayes as well as simple ad hoc predictors previously used for repeated measures or longitudinal studies with non-detectable values. We illustrate using real exposure data from a group of animal feed industry workers, and using HIV RNA levels from individuals in an HIV cohort study. We also present results from simulation studies.
Reneé H. Moore, PhD, is an Assistant Professor of Biostatistics at the University of Pennsylvania, Perelman School of Medicine. She holds a primary appointment in the Department of Biostatistics and Epidemiology and due to her extensive collaborative efforts with the Center for Weight and Eating Disorders holds a secondary appointment in the Department of Psychiatry. She received her B.S. in Mathematics from Bennett College and her M.S. and Ph.D. in Biostatistics from Emory University. Dr. Moore considers herself a collaborative biostatistician whose current focus is in the areas of clinical trials (design, analysis, data coordinating centers) and obesity research. Her statistical methodology research involves estimating longitudinal data subject to a limit of detection.
|Date:||Thursday, September 15, 2011
|Time:||Social time -- 6:00 pm
Buffet -- 6:30 pm
Speaker -- 7:20 pm
South College Ave.
University of Delaware
Newark, DE Directions/Map
|Menu:||(Catered by Aramark, University of Delaware Food Services)|
|Cost:||$20 for members/others, students $5|
|Note:||Please be sure to RSVP by this date. Aramark requires an accurate count of the number of plates three days prior to an event, and we pay by the plate.|
Return to Delaware Chapter ASA home page.