Goodbye, p-value: practical Bayesian statistics to replace frequentist statistics [online video]

Authors

Sean Koebley, Virginia Commonwealth University

Document Type

Presentation

Original Publication Date

2021

Streaming Media

Comments

Part of "How-to Talks by Postdocs," a series of instructional brown-bag lunch talks for the general VCU health sciences community taught by postdocs. For more information, visit https://rampages.us/howtotalks

Date of Submission

April 2021

Abstract

We've all heard about the serious limitations of frequentist statistics: p-hacking, misinterpreted results, and unmet assumptions of normality, etc.. Bayesian statistics presents an alternative. In Bayesian statistics, using the powerful simplicity of Bayes' theorem and modern computational muscle, the credibility of candidate parameter values is robustly reallocated to be most consistent with the data. In this seminar, after a brief introduction to the fundamentals of Bayesian statistics, we will apply our knowledge to that workhorse of experimental design: a comparison between two groups. We will execute both a standard t-test and a Bayesian estimation to determine the difference between groups, then compare the results.

Rights

© The Author

Is Part Of

VCU How-To Talks by Postdocs