Thinking About Data (Ethically): Power, p-Hacking, and Reproducibility
Date and Time
April 18, 2018
03:00PM - 04:30PM EDT
Location
Waterhouse Room, Gordon Hall, 1st. Fl., 25 Shattuck Street
REGISTER
This is part of the Research, Rigor, and Reproducibility Series
Speaker: Richard Born, M. D., Professor of Neurobiology and Director of the PhD Neuroscience Program, Harvard Medical School. HMS Bio
Workshop Details: Statistics is a powerful set of tools for distilling knowledge from data. However, if these tools are used irresponsibly, they can lead us down the path of false positive science—that is, scientific claims that may, in the short term, garner publications and honors, but that ultimately fail to reproduce. While some failures are expected (“honest mistakes”), many more arise through bad statistical practices, such as “p-hacking” and under-powered experimental design. I will discuss sources of these bad practices, some related statistical pitfalls, and some possible remedies through healthier habits of mind.
Upcoming workshops in this series:
5/1/18: "Irreproducibility in Published Research" with Jeffrey F. Flier, Caroline Shields Walker Professor of Medicine, Harvard University Distinguished Service Professor and George Higginson Professor of Physiology and Medicine, HMS (former HMS Dean)
5/4/18: "Experimental Design in Research" with David Glass, Executive Director, Muscle Diseases, Novartis Institutes for Biomedical Research and Lecturer, Department of Cell Biology, HMS
This is part of the Research, Rigor, and Reproducibility Series
Speaker: Richard Born, M. D., Professor of Neurobiology and Director of the PhD Neuroscience Program, Harvard Medical School. HMS Bio
Workshop Details: Statistics is a powerful set of tools for distilling knowledge from data. However, if these tools are used irresponsibly, they can lead us down the path of false positive science—that is, scientific claims that may, in the short term, garner publications and honors, but that ultimately fail to reproduce. While some failures are expected (“honest mistakes”), many more arise through bad statistical practices, such as “p-hacking” and under-powered experimental design. I will discuss sources of these bad practices, some related statistical pitfalls, and some possible remedies through healthier habits of mind.
Upcoming workshops in this series:
5/1/18: "Irreproducibility in Published Research" with Jeffrey F. Flier, Caroline Shields Walker Professor of Medicine, Harvard University Distinguished Service Professor and George Higginson Professor of Physiology and Medicine, HMS (former HMS Dean)
5/4/18: "Experimental Design in Research" with David Glass, Executive Director, Muscle Diseases, Novartis Institutes for Biomedical Research and Lecturer, Department of Cell Biology, HMS