The Substitute For P-values paper is popular. Received an email from the American Statistical Association informing me of the unusual viewing activity. The email copies this earlier email (I’m cutting out the names):
No problem! I also wanted to let you know of another article that appeared as one of “Taylor & Francis’ top ten Altmetrics articles” last week (and is still doing well). It’s “The Substitute for p-Values,” by William M. Briggs (Vol 112, Issue 519 of JASA). So far, it’s seen 149 tweets from 143 users, with an upper bound of 150,371 followers! Below is the Altmetric score:
All the best,
I had never heard of Altmetric, but on looking at their list of the top 100 papers of 2015, paper number 100 had a score of 854 (top had 2782). Fame still awaits.
Paper 100, incidentally, was “Human language reveals a universal positivity bias.” Not at this blog, buster.
The main email said this:
Dear Dr. Briggs, I just thought I would make you aware that your comment “The Substitute for p-Values” (http://amstat.tandfonline.com/doi/full/10.1080/01621459.2017.1311264)
Has been viewed more than 3,000 times and is still very popular on social media (see below).
Thank you so much for your contribution to JASA! [E], ASA Journals Manager
The link to the official paper is above (here too). The original post about it is here. The book page for Uncertainty, which contains all the meat and proofs of contentions in the paper, is here. Uncertainty can be bought here.
Don’t miss the free Data Science course, which puts all the ideas of the paper into action. This course is neither frequentist nor Bayesian nor machine learning/artificial intelligence, but pure probability.
Just look at that! The editors “best books” next to readers’ favorite book. The p-value measuring this correlation must be mighty wee! Weer than wee! Wee wee. All the way home!
Briggs says: “Paper 100, incidentally, was “Human language reveals a universal positivity bias.” Not at this blog, buster.”
“Not on this blog”
Then he posts a link to his anti-p-value paper, which very early on contains this:
“There are no good reasons nor good ways to use p-values. They should be retired forthwith. The reasons for this are many, and most are well explicated in McShane and Gal’s fine article. It is well to complain about something broken; it is better to provide a solution. This I will do, while highlighting philosophical deficiencies and fallacies of p-value-driven NHST.”
Did you see it: “…It is well to complain about something broken; it is better to provide a solution. This I will do,…”
Read closely: “…it is better to provide a solution. This I will do,…”
That is a “positivity bias.” And it IS here on this blog.
C’est la vie.
I looked through that Altmetric list of the top 100 papers and was surprised to count only 5 on the subject of climate change.
Congratulations such metrics actually reflect something. ( whether it’s something we sought after, approve of, or will have permanent relevance is a different issue).
This takes its rightful place in the pantheon of related postings by Gelman on ‘abandoning statistical significance’ and Harry Crane’s full-throated takedown of the Benjamin st al. paper advocating P < .005
Thanks for your Warriorship!
After that glowing praise I’m going to buy the book. How do I get it signed by the author?
“Positivity bias” is an embarrassingly science-y term and I’m sure glad I didn’t come up with it. It apparently refers to “A pervasive tendency for people, especially those with high self-esteem, to rate positive traits as being more true of themselves than negative traits.”
Before I looked it up, I thought the term had something to do with “positivism,” which denies the existence of the meta-physical, or maybe with the legal theory of “positive law,” which denies the existence of natural law and roughly means “law that is valid by the will of whoever made it.”
I’m not going to read “Human language reveals a universal positivity bias.” But I’ll bet dollars to donuts that there are p-values lurking somewhere within its analysis of human languages.
Because how could anyone do real science and find important Findings like “we tend to think very highly of ourselves,” without p values? Huh? Huh?
Good question. The XYL suggested bookplates, signed and which you could glue into the book. That sound good?