iiinova — where your digital self keeps evolving

Ben Bolker @bbolker.bsky.social · 01/10/2026

(maybe a little *too* dominated by the pandemic dip on this scale ...)

Ben Bolker @bbolker.bsky.social · 01/10/2026

github.com/bbolker/bbmi... (lazy-coded with help from Claude)

plot of monthly visits to US from Mexico/Canada/UK/Japan/France, log scale. Giant pandemic drop. Only Canada and France are trending down recently

Ben Bolker @bbolker.bsky.social · 01/10/2026

are the data easy to get? I always like to see these things on a log scale ...

200

Yup, that's like these from Agresti. (I should try the Agresti-Coull Beta(2,2) prior ...) I could try again with smaller baseline prob ...

from Agresti and Coull Am Stat 1998: figures of binomial coverage probability for n=5 (top row), 10 (bottom row) for Wald, Agresti-Coull ("Adjusted Wald"), Clopper-WIlson ("Exact")

Figure 1.3 from Agresti _Categorical Data Analysis_: Clopper-Pearson, Score, Wald binomial confidence interval coverage for pi = 0-1.

100

Claude derived it (!!) but it is mentioned in passing in Greenland et al. 2015 doi.org/10.1002/sim.... ... (they argue for more strongly regularizing priors than the Firth/t correction ...)

doi.org

041

Did you note the last figure, which includes 1-sided coverage for both tails? (Maybe you did & are just saying this should be highlighted more?) I started working on this after a Q from someone who was being hassled by a reviewer for using Wald tests -- I was surprised at how adequate they often are

200

Wald vs likelihood ratio tests: examples

Stats explorations w/ Claude remind me of "a 4-wheel drive vehicle just gets you stuck in deeper mud". Going down rabbit holes I would previously have avoided ... surprisingly (to me) small/complicated differences between LRT & Wald for coverage/type-1 error ... bbolker.github.io/bbmisc/wald_...

bbolker.github.io

3364

Ben Bolker @bbolker.bsky.social · 26/09/2026

At Cambridge University only senior fellows of colleges are allowed on the grass, or people accompanying senior fellows (at least this was the rule in the 1990s). Since the ducks go anywhere they damned well please I always said you should be allowed on the grass if you were accompanying a duck ...

Ben Bolker @bbolker.bsky.social · 24/09/2026

a lot of scaling to do here ...

The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher. During that time, the chips can process short queries so Google’s A.I., known as Gemini, can respond.

150

Ben Bolker @bbolker.bsky.social · 14/09/2026

Not so at my institution; I get lot of "oops, I'll be a few minutes late to the meeting, I have to update Zoom :-(" e-mails. This may be an option that the site administrator sets.

111

Ben Bolker @bbolker.bsky.social · 14/09/2026

github.com/jennybc/code... (recommended!) cites * Beautiful hedgehog cake: www.bbcgoodfood.com/recipes/hedg... * Homely hedgehog cake: www.reddit.com/r/funny/comm... i.imgur.com/peilfAh.jpg * Photos originally found at www.boredpanda.com/funny-cake-f...

github.com

GitHub - jennybc/code-smells-and-feels: Talk on code smells and feels and how to change that via refactoring

Talk on code smells and feels and how to change that via refactoring - jennybc/code-smells-and-feels

180

Ben Bolker @bbolker.bsky.social · 11/09/2026

Has an LLM found a new bug in TeX yet?

030

Ben Bolker @bbolker.bsky.social · 10/09/2026

Ironically, Claude refused to answer this question for me: "are there well-attested examples of dual-use (civilian/military) biological experiments on pathogens that selected on virulence, transmissibility, etc. to develop bioweapons capability in a vector-borne pathogen?"

140

Ben Bolker @bbolker.bsky.social · 10/09/2026

Also a little surprising that someone would work on a vector-borne pathogen if they were trying to develop a bioweapon? Maybe I'm missing something. (Although it sounds the associated hinkiness - routing to avoid access controls etc. - would be worthy of blocking in any case ...)

350

Ben Bolker @bbolker.bsky.social · 31/08/2026

we can haz preprint ... ??? (or not yet?)

Ben Bolker @bbolker.bsky.social · 26/08/2026

Probably an eastern black walnut (_Juglans nigra_)? www.inaturalist.org/observations... (c) Sandy Wolkenberg – some rights reserved (CC BY)

five eastern black walnuts, in their outer casings, slightly decayed, sitting on a bed of brown leaves with some green leaves in the foreground

One of my important roles as a senior academic: telling/reminding super-smart junior people that (1) they may soon have the novel experience of not being the smartest one in the room; (2) when they *are* the smartest one in the room, they should still try not to be jerks

527920

(Would a kickstarter help?)

any chance of a new print run (and availability outside the UK) ... ??

160

I was automatically suspicious of the large effect size, but hadn't had a chance to dig in - reminded when the study was quoted on @cbcradiocanada.bsky.social's "What on Earth" program today (I think - transcript unavailable & I'm too lazy to go through it ... www.cbc.ca/listen/live-... )

cbc.ca

Ben Bolker @bbolker.bsky.social · 21/08/2026

Anyone dug into back-and-forth on Frank (2024) doi.org/10.1126/scie... (estimates an 8% increase in infant mortality due to WNS →insect pops → insecticide [!])? Critique by Marx-Stoelting et al: doi.org/10.1007/s002... . Frank's response: doi.org/10.1007/s002... Rejoinder: doi.org/10.1007/s002...

140

I spent a little while on the issues list of one of those repos trying to explain to people (some of whom were actually well-meaning) why it wasn't practical to include everything (e.g. effects of environmental humidity and temperature) in the model ... github.com/mrc-ide/covi...

github.com

Any modeling of temperature and humidity etc? · Issue #266 · mrc-ide/covid-sim

https://www.google.com/search?q=wuhan+temperature+by+month https://www.youtube.com/watch?v=FI5MqXPSRD8 https://papers.ssrn.com/sol3/Papers.cfm?abstract_id=3551767 A quick grep of the code seemed to...

130

Ben Bolker @bbolker.bsky.social · 14/08/2026

that was the one I meant -- thanks.

050

Ben Bolker @bbolker.bsky.social · 14/08/2026

can someone point me to the (somewhat recent?) skeet along the lines of "I fitted <super-fancy model>, now I'm going to boil it all down to a simple linear slope _post hoc_ and make that hypothesis test my main conclusion <sarcastic>" ? (The original skeet was wittier and more eloquent ...)

270

Ben Bolker @bbolker.bsky.social · 11/08/2026

PS -- Git itself is GPL too, which helps with transportability, but (moving away from archival concerns to more day-to-day stuff) I worry about getting attached to/locked in by GH features (issues, pages, CI/Actions, PRs, ...)

Ben Bolker @bbolker.bsky.social · 11/08/2026

A couple of things: (1) AIUI OSF isn't taking down content, it's just not going to accept new content; (2) as mentioned, I'm much more worried about enshittification than outright collapse (e.g. "from now on only paying customers have API access"). (Linux is protected by the GPL.)

Ben Bolker @bbolker.bsky.social · 11/08/2026

Not sure what "Zenodo obviously is taking it down" means, pls clarify? I'd worry most about future potential enshittification of GitHub. Deletion/privatization of repos is also possible (although Z repos can also be privatized - deletion requires review ... www.openaire.eu/zenodo-intro... )

openaire.eu

130

Ben Bolker @bbolker.bsky.social · 07/08/2026

CRAN submission will be closed from Aug 5, 2026 to Aug 19, 2026 cran.r-project.org

cran.r-project.org

The Comprehensive R Archive Network

Danielle Navarro @djnavarro.net · 06/08/2026

Reposted by Ben Bolker

The title is a little pointed, but I am very serious here. Academics, medical researchers, and other organisations should not collect data on birth sex. It is rarely relevant, and it almost always leads to privacy violations for transgender people. Please share. I would like people to read this one.

blog.djnavarro.net

Separating sex and gender is statistical malpractice – Notes from a data witch

Statisticians and academics in Australia should ignore the terrible advice of most LGBTIQA+ organisations and read this post instead. The 2020 ABS standard is almost never the correct way to collect d...

14236119

Ben Bolker @bbolker.bsky.social · 04/08/2026

Sorry, but is this bait? (Yes, it's an incorrect definition ... would need at least an "or something more extreme" to be correct, unless I'm missing something ...)

171

Ben Bolker @bbolker.bsky.social · 03/08/2026

Whereas I remember being told by our insurance company that we needed an insurance rider for our "dangerous breed" dog (♂ rotty/?? mix, unknown but I think probably cocker spaniel), 70 lbs, dumb as a brick, absolutely sweet and completely harmless ... [see my profile pic]

100

Reposted by Ben Bolker

Martyn Plummer @martynplummer.bsky.social · 03/08/2026

What's new in JAGS 5.0.0 part 1: Initial Values martynplummer.wordpress.com/2026/08/03/w...

034

Ben Bolker @bbolker.bsky.social · 03/08/2026

I'm not sure whether `ggdist` had this functionality when we wrote the comment in bsky.app/profile/bbol... (I don't think so, but we could have missed it ...)

Ben Bolker @bbolker.bsky.social · 02/08/2026

My thoughts (but maybe you were asking @clauswilke.com?): it's basically impossible to come up with a universal/one-size-fits-all summary of trend and confidence. (R(t) vs r(t); magnitudes vs levels of certainty; when do you censor for lack of certainty ... every stakeholder is different ...)

Ben Bolker @bbolker.bsky.social · 02/08/2026

Nice. There's a deeper issue w/ CIs for curves (e.g. epidemic trajectories), known for a while in functional data analysis: see e.g. Juul+ (2021) doi.org/10.1038/s415... . (Our commentary was rejected both by Nature Physics & PeerJ ...) math.mcmaster.ca/~bolker/misc... github.com/aligharouni/...

A curve showing an ensemble of trajectories ("number newly hospitalized") from a stochastic epidemic model (black lines), as well as central regions computed in various ways.

Figure 1. Comparison of 90% central regions computed with different functional boxplot methods,
using epidemic curve ensembles from Juul et al.. FBD = functional band distance; J = number of curves
used for centrality calculation. Curve with J = 2 computed via the roahd package (Ieva et al., 2019):
curve with J = 50 used our own implementation of the functional band distance algorithm described by
Juul et al., curve with Mahalanobis (5 features) used Mahalanobis distance on features of interest
including the peak value of incidence, the time at which the peak occurs, the initial growth rate, epidemic
duration, and final size of the epidemic.

180

Ben Bolker @bbolker.bsky.social · 01/08/2026

although there's this very entertaining story from a few days ago x.com/gro_tsen/sta... (someone posts an AI-generated proof of the Collatz conjecture - famous, very simple but long-standing question - that passes Lean, but turns out to have been exploiting a bug in Lean itself ...)

x.com

X - The Everything App / X

From breaking news and entertainment to sports and politics, get the full story with all the live commentary.

121

Reposted by Ben Bolker

Vianey Leos Barajas @vianeylb.bsky.social · 31/07/2026

I have a project for this AI for science postdoc application: schmidtfellows.utoronto.ca To work on videos of migrating fish (eg salmon), for classification, segmentation, uncertainty quantification. Don’t need to be AI expert, but demonstrate high potential. Reach out if interested.

schmidtfellows.utoronto.ca

179

Ben Bolker @bbolker.bsky.social · 30/07/2026

**V. loosely scoped question** talked to a student about connections between Bayes & freq, e.g.: empirical Bayes [state-space, mixed models, penalized splines]; regularization/regularizing priors; sampling vs optimization (Monte Carlo EM/MCMC vs MLE/MAP). Suggest books/blog posts/papers on this ...?

2101

(also, I now see that \theta appears *three* times here - also as the mean of the first distribution statement ...)

000

Looks good except for "vairance" and "nromal" :-)

100

I guess I would expand it out a bit -- explain that we have a contribution from the likelihood of form (...) and a prior (...) Seeing two "distributed as" statements in succession that appear to define the same parameter in different ways looks weird to me. Maybe I don't read enough Bayesian stats?

100

OK, thanks, makes sense (although I do still find the notation pretty confusing ...)

100

I might be being a bonehead/reading too quickly, but is the second line here (and in the analogous equation farther down) supposed to be "s ~ ..." ? (i.e., effectively a half-normal prior on s, but making it full-normal is harmless because we're squaring s anyway) ?

$an equation $\hat \theta | \theta ~ Normal(\theta, s^2), \theta ~ Normal(\mu, \tau^2)$$

100

her interview on CBC on Sunday seemed much more positive (diplomatic?), FWIW: www.cbc.ca/listen/live-... (as was @bretdevereaux.bsky.social's take: foreignpolicy.com/2026/07/23/o... )

cbc.ca

021

Not sure quite what you're looking for, but iNaturalist has a huge collection - you could target particular known-to-be spotted species, e.g. panthers www.inaturalist.org/observations... There's also an API api.inaturalist.org/v2/docs/ which you could use to (vibe-??) code something ...

part of a panel of photos from iNaturalist showing pictures of leopards (Panthera pardus) in a variety of habitats