Dr. Yinglei Lai
Professor, The George Washington University, Dept. of Statistics
Professional Website
Assessing the discovery reproducibility from a large-scale association analysis
Wednesday, Oct. 23, 3 PM, RRI 421
Abstract: Reproducibility plays essential roles in scientific research. Magnetic Resonance Imaging (MRI) and genomic/proteomic high-throughput technologies have been widely used in brain and health research. The Dice Similarity Coefficient (DSC) has been commonly used for assessing the reproducibility of discoveries in a large-scale association analysis. However, in the current assessment of reproducibility, there is a lack of efficiency in the use of all available samples. More importantly, there is a lack of consistency with the reported discoveries identified based on all available samples. We have developed a probabilistic framework to assess discovery reproducibility based on all available samples. In our results, we demonstrated the usefulness of our approach and its advantages over DSC. We identified the minimal sample size required to achieve a given reproducibility rate, which provides an informative guidance for planning large-scale association studies.
No comments:
Post a Comment