Richard,
Thank you very much, the link to the New England Journal of Medicine abstract is exactly what's needed. I highly recommend it's reading to anyone following this thread who wants to understand the clinical situation as background that led to the publication of the subject AP article. Clearly this is a very serious "near-miss" for statistical significance and justification for some equally serious modifications to the way statistical decision-making criteria are implemented within regulated industries and professions. Assessment of the clinical evidence of a therapeutic benefit I must leave to subject-matter experts, but as a technical specialist I find it compelling. That this near-miss actually happened speaks volumes and deserves a wide audience.
My thanks also to author Malcolm Ritter who responded to me privately and recommended contacting Dr Scott Solomon, lead author of the NEJM publication, whose story Malcolm told in his AP article. Thanks also to David Couper who encouraged me to contact Dr. Solomon directly, in spite of my personal reluctance. David, your advice was excellent and perhaps it will happen now that Richard's post ha resolved my dilemma.
Each of you has helped a lot. All of this discussion convinces me that this story has legs, and that this is the tip of the iceberg.
Tom
Thomas D. Sandry, PhD
Industrial Statistical Consultant, Retired
------------------------------
Thomas Sandry
------------------------------
Original Message:
Sent: 11-27-2019 11:11
From: Richard McNally
Subject: Sorry, wrong number: Statistical benchmark comes under fire
Here's a link to the abstract:
https://www.nejm.org/doi/full/10.1056/NEJMoa1908655
------------------------------
Richard McNally
Statistical Fellow
Covance
Original Message:
Sent: 11-26-2019 18:12
From: Thomas Sandry
Subject: Sorry, wrong number: Statistical benchmark comes under fire
Hello All,
I have emailed Malcolm Ritter, author of the subject AP article, requesting any additional information about the nature of the drug trial and its outcome which he could supply while maintaining the confidence and privacy of the sources of his story. I'll post whatever he's willing and able to share publicly.
Tom
Thomas D. Sandry, PhD
Industrial Statistical Consultant, Retired
------------------------------
Thomas Sandry
Original Message:
Sent: 11-26-2019 06:47
From: David Corliss
Subject: Sorry, wrong number: Statistical benchmark comes under fire
Sharing here what I wrote on LinkedIn, which is a good channel for sharing these concerns with a wider professional but non-statistical community.
A good article on the perils of p-values, written in a clear, accessible manner. This is a good article to share with non-statisticians who need to use statistical results - business managers, engineers, physicians and other medical technicians, attorneys and other legal professionals, and so on. The article is excellent, but I could wish it added one more thing more: P-Hacking. In the article, a study is deemed a failure because it just missed an arbitrary significance level of 0.05. One risk we find in these cases - something people should watch out for - is repeating experiments or other tests over and over until one, by chance, barely achieves the 0.05 level, which is unwisely regarded as a gold standard.
My colleague Eric Vance mentioned the xkcd cartoon on p-hacking (jelly beans and acne) as good explanation of p-hacking for people who aren't statistical experts.
------------------------------
David J Corliss, PhD
Director, Peace-Work www.peace-work.org
davidjcorliss@peace-work.org
Original Message:
Sent: 11-17-2019 18:20
From: Glen Colopy
Subject: Sorry, wrong number: Statistical benchmark comes under fire
A familiar type of story in the news, mentioning in the ASA.
They have a fun quote "inside the arcane world of statistics"!
I've always thought we were a pretty cool and mysterious lot.
Sorry, wrong number: Statistical benchmark comes under fire
By MALCOLM RITTER AP Science Writer
https://www.miamiherald.com/news/article237281119.html
------------------------------
Glen Wright Colopy
DPhil Oxon
Data Scientist at Cenduit LLC, Durham, NC
------------------------------