INFOFUSION: Utilization of Multi-Source Data
Twelfth Biennial CDC Symposium on Statistical Methods
April 7-8, 2010

 

 

Agenda | Program at a Glance | Short Courses | Instructor Biographies | Committees

 

Descriptions

Methodological Issues in Biosurveillance

This short course will discuss the strengths and limitations of statistical detection algorithms, with focus on the issues related to developing, evaluating, and implementing such algorithms in biosurveillance applications. Open questions to be discussed include: (1) the appropriate metric or metrics for assessing and comparing algorithmic performance; (2) the utility of comparing algorithmic performance on real versus simulated data; (3) how to best modify existing methods to account for seasonal and other systematic effects in biosurveillance data; (4) how to maximize algorithmic detection capabilities within tolerable false positive rates; and (5) understanding when an electronic biosurveillance system “adds value” compared to other methods.

Back to Top

Information Visualization

This short course will focus on ways to visualize multivariate and multi-source data when the objective of the analyst is to gain insights and information about the underlying data processes. A secondary goal of the course is to discuss techniques to visualize the results in ways that help convey the concepts to non-statisticians. Potential topics for the course include techniques for visualization, such as scatter plot matrices and smoothing, parallel coordinate plots, Andrews’ curves and images, choropleth maps, and more. We will also look at ways to find alternative views of the data and interesting structure via linear and nonlinear dimensionality reduction and data tours. Finally, the course will include a discussion of the software tools that one can use to implement the techniques presented in the course. Where possible, the examples used to illustrate the techniques will be data from the health sciences.

Back to Top