Michael Plank

Relevant meme from our blog post on the topic: www.the100.ci/2024/12/05/w...

Screenshot of the "Does that use a lot of energy?" online app

Hannah Ritchie has built a fun little tool where you can compare energy usage of various products and activities.

This is super helpful imho, because it's so hard to develop intuitions even just about the scales involved here.

hannahritchie.substack.com/p/does-that-...

The Conservatives are, after all, well placed to know a lot about this morass, since they introduced it. In 2012, the coalition government launched the Plan 2 system of student loans and raised university fees across Britain to £9,000 per annum. To put Plan 2 in simple terms, loan repayments were laid out via a seemingly innocuous series of calculations. The first to consider is the threshold at which repayments begin. If you left education with, say, £27,000 worth of debt, you would only start paying it back once you met a predetermined salary. On its face, this might not seem like a particularly onerous demand. “Low-earning” graduates would avoid being saddled with repayments before they were financially able to begin making them, while their “high earning” peers could start chipping away at their debt, and provide an income stream for the state.

As any of my fellow literature or history graduates will tell you, however, the devil is in the details. For one thing, the threshold at which someone becomes a high earner was never particularly high and, following years of inflation, is now preposterously low. Rachel Reeves’ announcement that the government are freezing the threshold at April 2026 levels (£29,385) for a further three years only makes this worse. The real living wage for London is currently calculated at £28,860, which means that any London-based graduate making just £40 more per month than the minimum needed to live there will automatically begin paying their debt. In real terms, this means practically any graduate in any form of full-time work will be paying as much as 9 per cent of their income to the state, and for a very, very long time. Worse still, the amount owed by those graduates below the threshold does not remain static – it accrues interest, year on year, whether you’re working for low wages, volunteering, taking a career break or on maternity leave, ensuring that if you do pass the threshold some time later, you will be returning to find your original £27,000 much enlarged.

If the state’s attitude to what constitutes “high earnings” makes you think it’s oblivious to the concept of inflation, let me put your mind at ease. When it comes to the calculation of student loan interest, they are very conscious of inflation indeed. Each year, the interest charged on student loans is calculated by two components. The first is the Retail Price Index (RPI), which generally records a higher number than the Consumer Price Index (CPI). Governments prefer the latter, lower figure for many of their other calculations, just not when it comes to adding extra debt to every graduate in the country. To this is added a second component, a percentage tied to each graduate’s earnings, meaning that as your salary increases so too does the interest you’re paying on the loan you took out. If you think this seems like a predatory and punitive way to bilk students for as much money, and over as long a period of time, as possible, then you’re just about up to speed on this scandal, which amounts to a regressive stealth tax on every graduate in the UK. One which, it’s calculated, you would need to be earning £66,000 per year to pay off in anything like a timely fashion.

The debt burden of UK students is one of those things where, the more you look into the details, the more insane and predatory it is. So I tried my best to explain the numbers involved without making my, or your, head explode.

Suppose you know that x+y=100 and want to know what x and y are. You’re stuck right – there’s a whole range of possibilities for two numbers that sum to 100 (in maths we call this non-identifiability).

How well did this year's flu shot work?

Interim results: Decently, not perfect.

"the risk of medically-attended influenza A(H3N2) illness was reduced by 40% among vaccinated relative to unvaccinated individuals."

tinyurl.com/emw36bfa by Separovic et al., in @eurosurveillance.org

We'd be in a better place if university leaders would constantly beat this drum. Teaching critical thinking makes for better neighbors and citizens, while creating knowledge for knowledge's sake has driven massive economic gains, often in unpredictable ways. Interfering with this is folly.

Thanks Tyler Cassidy for leading the charge on this, the excellent team of co-authors, and to @matrix-inst.bsky.social for hosting us.

A Nonparametric Approach to Practical Identifiability of Nonlinear Mixed Effects Models - Bulletin of Mathematical Biology Mathematical modelling is a widely used approach to understand and interpret clinical trial data. This modelling typically involves fitting mechanistic mathematical models to data from individual tria...

Well it turns out if you know something about the distributions of x and y, you *can* (sometimes) separately estimate their population means.

We explored some biological applications of this idea here:
doi.org/10.1007/s115...

Now suppose you don’t just have one measurement of x+y but many, clustered around 100 but with some variation.

How does this help? Surely x=y=50 is just as likely as x=10 and y=90?

Suppose you know that x+y=100 and want to know what x and y are. You’re stuck right – there’s a whole range of possibilities for two numbers that sum to 100 (in maths we call this non-identifiability).

Go on tell us how many life minutes you are losing each year

😮

We're probably getting pretty close to 700 science jobs lost over this term of government, directly linked to reforms/cuts. And believe me: the country wasn't overflowing with them to start with.

Introducing CAPphrase (Comparative and Absolute Probability phrase dataset), an open access dataset containing over 150,000 probability-based language judgements: adamkucharski.github.io/CAPphrase/

This recent RCT of an "AI stethoscope" claims the technology "shows promise" for diagnosing cardiovascular conditions.

It does not.

It is a textbook example of the risks of conducting unprincipled 'per protocol analyses'. Once again, peer review at a major medical journal has failed.

🧵 1/

There's some amazing work here - an incredible breadth of projects from drug discovery & clinical trials, to avian influenza surveillance, TB control, and community-led initiatives for drinking water safety. Very proud to have contributed in a small way to this
www.teniwha.com/news/ebook

Bayesian joint modelling using @mc-stan.org of wastewater and hospital admissions in the US led by @kaitejohnson9.bsky.social and others at the US CDC

Join us for the launch event! We are excited to invite you to the official launch of the Global Society for Infectious Disease Dynamics (GSIDD) on February 26th, 2026.This launch marks an important milestone for our global communi...

Launch of the Global Society of Infectious Disease Dynamics. 🎉

2 online webinars on Thursday 26th.
Do join to see what it's all about, and get involved in the community as it kicks off!

www.gsidd.org/post/join-us...

Ah the animation's not working, maybe this will work
images.thespinoff.co.nz/1/2020/03/Co...

You might be interested in this by the excellent @siouxsiew.bsky.social and @xtotl.bsky.social (not for vaccines but similar idea)

It drives me crazy that most people think that COVID-19 vaccines don't reduce transmission. They don't bring Reff <1, but the difference between spreading an infection to 2 people vs 1 person is huge at the scale of the population.

There's some amazing work here - an incredible breadth of projects from drug discovery & clinical trials, to avian influenza surveillance, TB control, and community-led initiatives for drinking water safety. Very proud to have contributed in a small way to this
www.teniwha.com/news/ebook

Except literally every data analysis or model I've ever asked AI to build has contained at least one error: www.nature.com/articles/d41...

Common response: "But someone just needs to give AI detailed instructions..."

But who is the "someone" in that sentence? And how did they get their expertise?

The last part abt how measles disrupts immunity vs other infections is, imo, underappreciated (& mechanism was recently clarified). It's part of why measles vaccination was seen to help beyond just reducing measles complications - not getting measles reduces severity of subsequent infections too.

Great talk on composable infectious disease modelling. Making real-time outbreak modelling more principled and flexible has been a long-standing challenge, and Sam has been leading the charge on solving it. With a growing group of people keen on this it feels like we're finally getting somewhere.

What could possibly go wrong?

"With user-pleasing AI agents getting faster at tasks – and potentially subject to less user scrutiny – there’s more need for statistical thinking than ever"

I think news stories like this have been so frequent over the last six years that it's now "common knowledge" that covid ages your internal organs.

But, it's simply not true.

1/

Rakiura is a beautiful place - enjoy!

An example of how impacts of interventions can differ at individual and population levels, and are conditional on other events. Given a car hits you, the helmet helps. But mandating helmets could lower the # of bikes on the road & thus might increase the chance of getting hit in the first place.