|
|
|
|
Title: |
My RA in |
|
Speakers: |
Bryan Shepherd, Graduate student, Biostatistics |
|
Abstract: |
I spent this past summer in |
|
|
|
|
Title: |
Network meta-analysis for indirect treatment comparisons |
|
Speakers: |
Thomas Lumley, Assistant Professor, Biostatistics |
|
Abstract: |
I present methods for assessing the relative effectiveness of two treatments when they have not been compared directly in a randomized trial but have each been compared to other treatments. These network metanalysis techniques allow estimation of both heterogeneity in the effect of any given treatment and inconsistency ("incoherence") in the evidence from different pairs of treatments. A simple estimation procedure using linear mixed models is given, and used in a meta-analysis of treatments for acute myocardial infarction. |
|
|
|
|
Title: |
Spatio-temporal Bayesian hierarchical
models for mapping disease rates |
|
Speaker: |
Erin Conlon, Biostatistics |
|
Abstract: |
Hierarchical models provide a
mechanism for stabilizing disease rate estimates from small areas, while
maintaining geographic resolution. Spatial correlation among rates is
possible, due to unobserved covariates such as environmental or
socio-economic effects. Spatial variation is incorporated into the
hierarchical model through random effects with a spatial prior structure. In
such a model, regional rate estimates are stabilized by combining information
from neighboring regions. In many past applications, the neighborhood
definition was fixed a priori as regions sharing a common boundary. We use
two methods for generalizing the neighborhood structure that allow the data
to inform on the strength and geographic extent of spatial similarity. We
compare posterior parameter estimates and model fit for the various
approaches using 21 years of lung cancer mortality data in counties in the
state of |
|
|
|
|
Title: |
What if the number of 'things' we don't know is one of the
things we don't know? |
|
Speaker: |
Chuan Zhou, Graduate student, Biostatistics |
|
Abstract: |
Recent advances in technology have generated enormous data sets such as satellite data, climate data, web-site visiting information and, of course, microarray data. Often the volume of these data sets prevent us from extracting information directly. Cluster analysis is now recognized by many researchers as an important and useful tool for data mining. However, inferring the number of clusters (components) underlying the data remains a practical and theoretically challenging problem. This is more often a scientific question rather than a statistical question. In this talk, I will give a survey of the various ways to estimate the number of clusters in "classical" clustering and model-based clustering frameworks. Discussion and feedbacks are most welcome. |
|
|
|
|
Title: |
Aspects of cancer
biomarker development |
|
Speaker: |
Margaret Pepe, Professor, Biostatistics |
|
Abstract: |
Recent advances in biotechnology, including expression-array methodology, proteomics and immunology, promise to yield biomarkers for early detection of cancer. Population cancer screening based on such biomarkers could greatly impact on the mortality and morbidity associated with cancer. Biomarker development is a relatively new field for the research community. Guidelines are needed to guide the development process. We propose that biomarker development be structured into 5 distinct consecutive phases of research: Phase 1, the initial exploratory phase to identify potentially promising markers; Phase 2, the clinical assay development and validation phase, using clinical specimens from cases with established disease and suitable controls; Phase 3, the phase that determines if the biomarker detects disease early, before it becomes clinically evident, using specimens from clinical repositories; Phase 4, the initial prospective population screening phase that establishes the number and nature of cancers detected; and Phase 5, the cancer-control phase, that determines the net benefits and costs of the biomarker based population cancer screening program. For each phase, we outline objectives and key elements of study design. Rigorous and efficient development of useful biomarkers is likely to be facilitated by a structured approach, as is already in place for therapeutic drug development. Criteria are now needed for determining when a biomarker has successfully completed one phase and can proceed to the next. |
Last Modification: