James Steele

Some @ausrepro.bsky.social @aimosinc.bsky.social updates 🧵1/3
Check out the recording of previous webinar with Prof Jennifer Byrne & Prof Ben Mol "Understanding threats to reproducibility in published research: Paper mills, research fabrication & more"
youtu.be/u8A6wwnHUDI
@jabyrnesci.bsky.social

Here it is osf.io/w2qpx

Plus the hot take that, at least above RDA (and based on reanalysis of the Nunes et al. 2022 data), protein intake barely matters for lean mass/muscle adaptation.

Finally venturing properly into the fun world of Bayesian model averaging. We're preparing a pre-reg for updating a systematic review and conducting dose-response meta-analyses... there's several plausible candidate functions, and probably more, to explore so it's a very good use-case.

Can't remember if it was Fisher, Neyman or Pearson but I think the original framing was it was the hypothesis that the researcher attempted to "nullify". I like Meehl's framing of the strong use too in NHST... The null should be set at what the theory predicts and that tested (nullified).

🔥🔥🔥

There's a lovely irony to this 😉

How to add packages to a .bib file from within a .qmd file How to add packages to a .bib file from within a .qmd file - citations.qmd

Okay, came upon @datavisfriendly.bsky.social 's little golden nugget here 💰:

github.com/quarto-dev/q...

And just had to expanded on it - so I give you:

***How to add packages to a .bib file from within a .qmd file***

gist.github.com/mattansb/6f9...

#rstats #quarto

{targets} goes in the list for me now as I feel in the past couple years I've fully embraced the approach.

Robert been posting some cool snippets and bits from it over on LinkedIn... Definitely gonna have to get a copy.

Interested in the latest metascience? Come to Eindhoven October 17th to listen to Dr. Duygu Uygun-Tunç (University of Chicago) and Dr. Lisa Spitzer (Leibniz-Institut für Psychologie) as keynote speakers, and Dr. Aaron Peikert for a workshop collaboration through Git! All free!

Still trying to convince folks in exercise science that's the rough shape of adaptation in pretty much everything with exposure over time. But those bro's still wanna hold out hope that they can grow to infinity (hopefully they don't realise that the log is approx but still technically monotonic 😂)

Octopus | Built for Researchers Get started with Octopus, an open access platform where researchers can read, review and publish findings at all stages of the scientific process.

Have you seen www.octopus.ac?

It's a platform for publishing separately and combining the various components of research programmes.

@octopus-ac.bsky.social

@cik.bsky.social

A good example is our recent work on Bayesian approaches to determining whether additional research might be wasteful... We use a specific example area of literature, and it could have just been a metascience project highlighting the waste there... But we wanted it to be mostly an educational piece.

I was maybe a bit derisive of the typical sort of low hanging metascience fruit research the other day... I do see a place for it. I definitely have the position personally that if I am doing it though it must serve the purpose of providing a case-study on how to actively do things better...

I've been reflecting on my contributions and approach to "Metascience"... I think I can best sum it up as trying to workout how to do my own substantive work better at all levels and then, where I figure something out or find an approach, communicating that to others...

If your name is on a paper you didn't contribute directly to, or even carefully read one time, you are a dishonest fraud, end of. If you publish as part of a cabal that puts everyone's name on every paper regardless of direct input, you are all dishonest frauds.

PSE8 Visit the post for more.

Do you want to go to a metascience conference where AI will not solve all problems, you can ask critical questions, and El$evier is not invited? Sign up for the mailing list for the 8th Perspectives on Scientific Error conference in 2026 in Leiden! perspectivesonscientificerror2026.wordpress.com

And later work on this e.g. meehl.umn.edu/sites/meehl.... and meehl.umn.edu/sites/meehl....

screenshot of an email to editors, that is available in full in the link in the post

I love Registered Reports because it's no fun wrestling with reviewers about design decisions that we can do nothing about after data collection. But still, SO few journals in my research area accept them.

So I just fired off a load of this email. Here if you want to do the same: osf.io/n6bx9

Tbf I've seen it even in the case of experimental designs... Someone over on LinkedIn claiming it can be used in A/B testing for it. Absent intervention replication it's just noise mining though.

Is it just me or do I keep seeing part of the grift as promising "precision" intervention identification i.e. true treatment effect heterogeneity, when thrown at designs just incapable of determining that in the first place?

#statsky #episky

Oooh this is great... At first glance feels like it provides a causal inf formalism to elements of the ontologies of folks like Bhaskar/Cartwright, even Dennett/Hofstadter, in terms of causally efficacious phenomena arising out of complexity. Adding to the must read list.

We finally extended robust Bayesian meta-analysis to multilevel settings. Now, you can fit 3-level publication bias-adjusted model-averaged meta-regression models in R (and in about a week in @jaspstats.bsky.social too!)

Casual (maybe juvenile trapdoor) spider strolling across the path on this morning's walk.

Meh, meant SIPS 2023 (as my phone's photo memories just reminded me 😅).

As someone who is ostensibly not a psychologist but attended my only SIPS in 2022, and attended Metascience 2025, I have to say this echoes my feelings too... even after SIPS in 2022 I commented that it had ruined me for other conferences, and I felt that the reinforced the first day of Metascience.

Oh you're in for a treat... It's a British institution at this point. And so many historical seasons to watch.

That acts as the constraining influence, then the posterior estimate from a given study can be contrasted with the prediction from theory or theories. That sounds more aking to what you suggest in terms of priors (and of course checking sensitivity).