Lightnews — Scholar-powered news

Adam Kucharski @adamjkucharski.bsky.social · 1d

It's a great metric if, for example, you happen to have an extremely well cited single author paper (as this will get counted three times...)

1 10

Adam Kucharski @adamjkucharski.bsky.social · 1d

2. The C-score was created by John Ioannidis and colleagues.

3. Over the years, the C-score has often placed Ioannidis in the top 20 scientists worldwide.

3 2 13

Adam Kucharski @adamjkucharski.bsky.social · 1d

Three things that are probably worth knowing about the 'Stanford top 2% scientists' metric that people are sharing.

1. It's based on a rather subjective 'C-score', where:

5 6 16

Adam Kucharski @adamjkucharski.bsky.social · 1d

Would be interested how it coded that up, but I suspect it did it on discretised 1 day intervals, which would propagate some error over time (e.g. wrong peak timing etc.)

1 1

Adam Kucharski @adamjkucharski.bsky.social · 1d

In discussions of AI for epidemic modelling, important to remember that a simple SIR model is the equivalent of print("Hello World") – a useful introductory step, but only the very start of the work required for a good model.

1 2 27

Reposted by Adam Kucharski

Jonn Elledge @jonnelledge.bsky.social · 3d

An experiment:

If you see this post, please like it. Am comparing numbers to the other place for a piece.

69 140 7K

Adam Kucharski @adamjkucharski.bsky.social · 2d

Now he’s wrapped in up in a more flexible package {overshiny}, which makes use of Epiverse {epidemics} package and its ability to automatically handle complex social mixing and demographic structure, which in turn is powered by {odin}, allowing users to customise models if needed.

Try it out here:

Tutorial: Overlays for interventions in epidemic modelling

nicholasdavies.github.io

3 11

Adam Kucharski @adamjkucharski.bsky.social · 2d

Want to build an interactive dashboard so others can explore epidemic scenarios? For COVID, @ngdavies.bsky.social spearheaded a great drag-and-drop approach, which made use of the JavaScript-based nature of RShiny... 1/

2 10 31

Adam Kucharski @adamjkucharski.bsky.social · 9d

That was the opening question in the panel at the end!

1 2

Adam Kucharski @adamjkucharski.bsky.social · 9d

It's an important distinction, but wouldn't really fall under discussion of drivers of misinformation (i.e. people believing things that are demonstrably false)?

1

Adam Kucharski @adamjkucharski.bsky.social · 9d

At our recent @lshtm.bsky.social event on misinformation, Chris Whitty gave some useful reflections on common causes and some important considerations for handling it.

Full event: www.youtube.com/live/H_nClQ2...

3 21 54

Reposted by Adam Kucharski

Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 9d

Was asked about collinearity again, so here's Vahove's 2019 post on why it isn't a problem that needs a solution. Design the model(s) to answer a formal question and free your mind janhove.github.io/posts/2019-0...

tl;dr

Collinearity is a form of lack of information that is appropriately reflected in the output of your statistical model.
When collinearity is associated with interpretational difficulties, these difficulties aren’t caused by the collinearity itself. Rather, they reveal that the model was poorly specified (in that it answers a question different to the one of interest), that the analyst overly focuses on significance rather than estimates and the uncertainty about them or that the analyst took a mental shortcut in interpreting the model that could’ve also led them astray in the absence of collinearity.
If you do decide to “deal with” collinearity, make sure you can still answer the question of interest.

3 33 110

Adam Kucharski @adamjkucharski.bsky.social · 10d

It's a wonderful historical quirk that a crucial improvement to the greedy algorithm came from Hungary.

1 1 13

Adam Kucharski @adamjkucharski.bsky.social · 13d

I wonder how many people who believed, despite all the available evidence, that vaccines cause autism have now happily switched to believe, despite all the available evidence, that Tylenol/paracetamol causes autism.

Reminds me of Kevin Simler’s piece on ‘crony beliefs’:

meltingasphalt.com

3 2 32

Adam Kucharski @adamjkucharski.bsky.social · 13d

“My mummy fights fires. Not big fires like a firewoman. Just metaphorical fires.”

My son getting to grips with parents’ jobs.

2 4 34

Adam Kucharski @adamjkucharski.bsky.social · 13d

I believe so!

1

Adam Kucharski @adamjkucharski.bsky.social · 14d

I'm giving a free public talk asking 'How certain is certain enough?' at @lshtm.bsky.social on Monday 24th November at 5.30pm, as part of the Global Health Lecture Series. From epidemic response and clinical trials to online misinformation and AI, I'll look at how truth emerges – and why it falters.

How certain is certain enough? | LSHTM

When faced with a new health threat, there are three crucial questions we must consider. What is going on? What can we do about it? And how can we communicate these things effectively? This talk will

www.lshtm.ac.uk

1 17 42

Reposted by Adam Kucharski

Martin Humphries @martinhumphries.bsky.social · 15d

Nearly finished Adam Kucharski (@adamjkucharski.bsky.social)'s book 'Proof', and came to this fascinating quote

4 15 47

Reposted by Adam Kucharski

David Hodgson @dchodge.bsky.social · 17d

🚨 New paper out in PLOS Computational Biology! 🚨

We're excited to share our new paper, serojump, a new probabilistic framework and R package for inferring infections and antibody kinetics from longitudinal serological data.

📄 Full paper: tinyurl.com/re7du3t2
R package: seroanalytics.org/serojump

A serological inference package using reversible jump mcmc

The `serojump` package provides tools for fitting serological models to antibody kinetics data using reversible-jump Markov Chain Monte Carlo (RJ-MCMC). It enables researchers to model the dynamics of...

seroanalytics.org

2 5 18

Adam Kucharski @adamjkucharski.bsky.social · 18d

Lesson in the importance of early file management choices – many of my COVID projects are still in a '2019-nCoV' directory...

1 19

Reposted by Adam Kucharski

David Hodgson @dchodge.bsky.social · 21d

Just launched an interactive Bayesian epidemic modelling platform that runs entirely in your browser!

No downloads, no installations, no expensive software licenses. Just open the link and start modelling disease dynamics with real-time parameter estimation.

>> widget-bayesian-sir.davidhodgson.me

Interactive Bayesian Epidemic Modelling

widget-bayesian-sir.davidhodgson.me

2 6 22

Adam Kucharski @adamjkucharski.bsky.social · 19d

Please, no more papers with cartoon 'team members' to anthropomorphise slightly different prompts to the same LLM.

1 3 43

Adam Kucharski @adamjkucharski.bsky.social · 22d

Easy version of 9-9-6: Do 12 hours work (or ‘work’), 6 days per week.

Hard version: Your young children only go to bed at 9pm, then they wake up 9 times between them during the night, then are up and ready to go at 6am.

1 17

Adam Kucharski @adamjkucharski.bsky.social · 23d

Just over an hour a way! 👇

Adam Kucharski @adamjkucharski.bsky.social · 23d

Looking forward to chatting with Eugenia Cheng about maths, proof and our notions of what 'equal' means live at 1pm ET/6pm UK time today: open.substack.com/live-stream/...

LIVE SOON: Discussion with Eugenia Cheng on maths, equality and proof

Starting Sep 17 at 1:00 PM EDT

open.substack.com

1 3

Adam Kucharski @adamjkucharski.bsky.social · 23d

Original piece: ourworldindata.org/much-better-...

The world is awful. The world is much better. The world can be much better.

It is wrong to think these three statements contradict each other. We need to see that they are all true to see that a better world is possible.

ourworldindata.org

3 6