|
|
 |
|||
 |
 Overview | Research | Education | Resources | Partnerships | Home |
||
 |
|
----------------------------------------------------------------------------- NIH BCIG Speaker Event:
"What's the Question: Statistical Design and Methods for Proteomics, Microarrays,
and Other High Dimensional Data" NIH BCIG Speaker Event:
"What's the Question: Statistical
Design and Methods for Proteomics, Microarrays, and Other High Dimensional
Data" Summary: Scientists have been excited about the advent of high dimensional tools to quickly screen for genes and proteins of interest. Unfortunately, spurious findings from experiments using these laboratory methods have become the rule rather than the exception. Because of the large number of outcomes being investigated simultaneously, statistical and computing methods had to be re-evaluated. The methods are similar to those used in a "standard" simple clinical trial, but since the number of outcomes far exceeds the number of samples, attention to detail in the experiment and the data is critical. Every study answers a question and has hypotheses, even if the study is exploratory. To help define the correct statistical methods and sample size a study can be classified in many ways. During the presentation we will focus on three general questions: 1) Class comparison, is this gene differentially expressed; 2) Class prediction, building models using multiple features so a new sample will be classified as belonging to a certain class; and 3) Class discovery, closely related to data mining. These three questions help define the experimental design from a statistical point of view. For each of these questions we will review typical tests and analysis techniques and describe what problems can occur. We will explore what computing methods can expose and more importantly hide in the data. Good study design up front can later provide a frame work to help tease out non-biological artifacts present in the data; by teasing out these artifacts and using good design we lessen the chance of spurious results. Bio Laura Lee Johnson graduated with a BA in mathematics from the University of Virginia and a Ph.D. in biostatistics from the University of Washington's School of Public Health and Community Medicine. She was a pre-doctoral trainee at the Northwest Veterans Affairs Health Services Research and Development Center of Excellence in Seattle and a Presidential Fellow in the Department of Biostatistics at the University of Washington where she taught biostatistics to clinical fellows. Currently she is a Research Fellow helping lead the biostatistics core in the Cancer Prevention Studies Branch in the Center for Cancer Research at the National Cancer Institute. Her research interests include analysis of biomarker data, statistical methodology for non-parametric joint modeling of longitudinal and survival data, and longitudinal crossover designs. johnslau@mail.nih.gov More information Substantive refreshments will be served.
"The Influence of Topology
on Artificial Dynamical Processes" "The Influence of Topology on Artificial
Dynamical Processes" Dr. Tomassini will be available to visit with faculty from June 23 until June 25. If you would be interested in meeting with Dr. Tomassini, please let Dr. Monty Kier know your preferred time and date at: kier@mail2.vcu.edu. The topological structure of artificial and natural systems has a marked influence on the dynamics of the processes that can take place. After a brief introduction, we explore artificial evolution as a case study. We present models for selection pressure in structured populations for regular lattices, random graphs and small-world graphs, and we discuss the role of the network structure on the results.
|
|
