Also this week: the US government vs. frequentist statistics, a survey on measuring excellence in scientific outreach, outlier sheep, the essential thought for anyone giving a talk, and more

**From Meg:**

Alex Bond had a post reflecting on what he’s learned in his 10 years in science. It’s spot on. (ht: Stephen Heard)

Sigh. There’s always one, isn’t there?:

ht: @McNameeJason

**From Jeremy:**

The EEB and Flow with a really interesting post on ecology’s first null model war (from the 1920s!)–and how it didn’t prevent the second.

Steven Frank with a new preprint linking the Price equation and Fisher’s fundamental theorem of natural selection to…[wait for it]…d’Alembert’s principle from classical physics (a generalization of Newton’s second law of motion, F=ma). I love this kind of stuff, identifying deep connections between seemingly-unrelated ideas. Here, the connection is that relative fitness is like relative motion (i.e. relative to a frame of reference). Interestingly, Frank seems (?) to be getting away from the idea that the Price equation has deep linkages to information theory, now preferring to think of the conservation of key quantities (e.g., total probability, mean relative fitness, the sum of direct and inertial forces) as what’s truly fundamental. (UPDATE: Just found this new preprint from Frank, linking his new d’Alembert-based perspective back to the information theory perspective. He does indeed now believe that conservation of total probability is the really fundamental thing here, and that MaxEnt is a “useful but sometimes unnatural” way to express the “geometric” constraints imposed by conservation of total probability.)

Advice on academic job hunting in the US for non-resident foreigners. I can’t vouch for it, but it sounds reasonable to me.

The Global Young Academy (an organization of top young scientists, with which my friend Rees Kassen is heavily involved) is conducting a survey on scientific engagement and outreach. They want to determine how universities and other scientific institutions measure and reward engagement and outreach, and how their employees and administrators *think* they measure and reward engagement and outreach. Click the link to take the survey; there are separate versions for profs/NGO scientists/administrators and students/postdocs.

Deborah Mayo and Andrew Gelman are rightly horrified by the way a US government webpage defines statistical significance and P-values.

Remember: when you give a talk, don’t be scared–you’re the smartest person in the room.

A rare retraction in ecology, due to an innocent mistake arising from the lead author’s serious illness. A very unusual and unfortunate situation; there’s no suggestion of misconduct or incompetence on the part of anyone involved.

Geez, the Google autocomplete suggestions for “my phd” are depressing.

Oh, that retraction story is really sad. It reminds me that I haven’t given as much thought as I should to this post of Stephen Heard’s on publication power of attorney: https://scientistseessquirrel.wordpress.com/2015/01/17/whats-a-publication-power-of-attorney-and-why-should-you-have-one/

nor to this post from Alex Bond on academic wills:

https://labandfield.wordpress.com/2013/07/14/making-a-scientific-will/

Not fun to think about, but important.

Pat Schloss wins the internet for today, referring to that sheep scatterplot:

[First comment, long-time reader :-)]

Thanks for the link to the interesting preprints by Frank. Lanczos’ book is one of my favourites and I always enjoy work combining probability, invariance principles and geometry (plus biology!). As noted in the other preprint (and, from what I’ve seen btw, by many others) this stuff shows up everywhere, especially mechanics, thermodynamics and statistical inference. It can be hard to tell these subjects apart sometimes!

I first came across these general ideas when I was an undergraduate majoring in continuum mechanics – it’s quite common to derive thermodynamically-constrained constitutive equations on the basis of a combination of d’Alembert’s principle and maximum entropy production. Hans Ziegler and Dominic Edelen both gave strong geometric interpretations to maximum entropy production in this context. After moving more into stochastic modelling and statistical inference I’ve also come to feel that conservation of probability (another invariants) and geometric ideas are more fundamental. There were/are a few attempts to explore these directions in the nonequilibrium/stochastic thermodynamics literature. And, in fact d’Alembert’s principle is itself strongly related to differential geometry (symplectic and contact geometry). Never actually written anything about it myself (or fully synthesised it all) though!

another invariants -> and other invariants

Interesting, and thanks for the pointers!