Activity Number:
|
8
- Statistics and the Reproducibility Crisis
|
Type:
|
Invited
|
Date/Time:
|
Sunday, July 30, 2017 : 2:00 PM to 3:50 PM
|
Sponsor:
|
The American Statistician
|
Abstract #321875
|
View Presentation
|
Title:
|
The Tools for Reproducibility Exist; P-Values Aren't the Problem; It's Time for the Real Work of Data Science
|
Author(s):
|
Jeffrey Leek*
|
Companies:
|
Johns Hopkins Bloomberg School of Public Health
|
Keywords:
|
p-values ;
reproducibility ;
data science ;
data science as a science ;
dsaas
|
Abstract:
|
There is a lot of discussion surrounding the "crisis" of reproducibility and p-values. This has been driven by difficulties in the cultural transition from data poor to data rich research. The source of the crisis has been attributed to lack of tools for reproducibility or p-values and hypothesis testing. In my talk I will discuss an alternative explanation: that data analysis is a human behavior and we understand it relatively poorly. I propose that we should study "data science as a science" and campaign for more statistics education across curricula rather than focusing on a specific tool or measure of significance.
|
Authors who are presenting talks have a * after their name.