Léo Varnet @leovarnet@qoto.org

Our preprint "Shifting the level of selection in science" should be an interesting read for anyone interested in recognition and rewards in science. We discuss how moving from rewarding individuals to rewarding teams could have important benefits for science.

I thought I'd remind you - our revision is stuck in review limbo due to the recent drama at Perspectives on Psychological Science, but it's a paper that I think deserves to be read as is.

Trahison, fausse humilité, mensonges, simplifications extrêmes et solitude de la majorité : voici le menu de notre émission @arretsurimages de la semaine qui décortique la communication du gouvernement autour de la #ReformeDesRetraites
Avec les analyses de @dan_mdpt et Arnaud Mercier
https://www.arretsurimages.net/emissions/arret-sur-images/retraites-la-simplification-a-outrance-de-la-reforme-est-une-trahison

via Judea Pearl:
Correction to #Bookofwhy: The concept of a controlled clinical trial was first described in 1662 by Jan Baptist van Helmont (not J. Lind, 1747) who wrote: "Let us take out of the Hospitals… 500 poor People, that have fevers or Pleuritis. Let us divide them in Halfes, let us cast lots, that one halfe of them may fall to my share, & the others to yours; I will cure them without blood-letting & sensible et you do, as ye know … we shall see how many Funerals both of us shall have…"

Multilayer perceptrons with 1 hidden layer approximate functions using sums of ridge functions. Approximation rates have been stated by Andrew Barron. http://www.stat.yale.edu/~arb4/publications_files/UniversalApproximationBoundsForSuperpositionsOfASigmoidalFunction.pdf

JASA

New cover! January features "Low dimensional measurement of vowels using machine perception." https://doi.org/10.1121/10.0016845

#acoustics #MachineLearning #speech

C'est marrant, le gars qui va faire une conférence pour l'AFIS sur le glyphosate a le même nom que celui qui racontait n'importe quoi sur le site de l'AFIS et sur France Culture à propos du glyphosate.

Ça ne peut pas être la même personne. Ils n'auraient pas osé ! … si😱
Thread

Google Trends stinks!

While frequently used by researcher Google Trends seems to be a highly unreliable source of historical trends, Alexandra Franzén argues based on an experiment revealing Google Trends inability to replicate its own results.

Compare for instance the two identical searches conducted at different times below.

"Big data, big problems: Why scientists should refrain from using Google Trends" is out now: https://doi.org/10.1177/00016993221151118

@sociology @dh @politicalscience @academicchatter

It has been a pleasure to see so many people find value in my paper on 'Sample Size Justification'. And yes, it is a long article! Did you know you can listen to me reading it out loud on #Soundcloud? You can put it on as you do chores in the house 😃🧹
https://on.soundcloud.com/VBB6q

Sampling a signal is equivalent to periodizing its Fourier transform. For non-smooth signals (with non-compact Fourier transform), it creates aliazing. https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem

NIH funding is not a meritocracy. Luck — both good and bad — plays a huge role, as does bias. This truth can be depressing to acknowledge but can also set you free (e.g., from being too hard on yourself).

Corollary: until we fix the system, submit a lot of (respectable, if not perfect) grants if you want to get funded.

#NIHgrants #ECRgrantAdvice

@JaneEyreZH

Personally, I set aside writing times in my calendar. When someone asks if I can meet then, the answer is "no, I'm sorry. It's already booked."

I have also long ago found my own personal "efficiencies" - I write best in the morning, so I reserve those times for writing.

There are also a whole slew of "tricks" I learned from my creative writing background - including

(1) the vomit draft - always edit on paper (electronic paper is OK), just don't edit in your head.

(2) the "Hemingway point" - when you've written a lot for the day, stop in the middle of a sentence. That way, when you come back tomorrow, you can start writing by finishing that sentence.

(3) make it habit - don't just write when the spirit moves you

(4) rewrite by retyping / rewriting, not by moving paragraphs around, even if you are retyping the same text over

(5) make sure there is time to edit. After you've written something, put it aside for a week or a month and come back to it. (Write other stuff in the meantime see note 3.) Then rewrite [note 4] that month-old text.

Importantly, there is no one "right way" to write. Try stuff out. Your mileage will vary. Find your own personal efficiencies. And yes, your efficiencies will change as your life changes. Be prepared for that, and then go find the new efficiencies.

Oh, and (6) Don't forget to READ! Read all kinds of things - fiction, not; in-field, out; great literature, trashy literature; great science, great art. Sometimes people get so wrapped up in their own writing, particularly as other life-needs occur, they don't do enough reading.

Conditional distributions defines a parametric collection of probability distributions computed from a joint distributions of two variables (here in the simple case of a density with respect to a product of measures). https://en.wikipedia.org/wiki/Conditional_probability_distribution