Tom Chivers Profile picture
Dec 9 2 tweets 1 min read
absolutely amazing. I'm writing about The Dress (it's Bayesian, you see) and this demonstration is almost uncanny.
image credit: Figure design by Kasuga~jawiki; vectorization by Editor at Large; "The dress" modification by Jahobr. Creative Commons license sr.wikipedia.org/wiki/%D0%A5%D0…

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Tom Chivers

Tom Chivers Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @TomChivers

Jan 10
on the continued obsession - including in official government advice! - with cleaning surfaces and washing hands to combat Covid, despite a near-total absence of evidence that it spreads that way unherd.com/thepost/repeat…
(since I wrote this I think the Liverpool FC training ground had a "deep clean" to interrupt the outbreak there, which almost certainly will not have helped)
A fair number of people are responding saying things like "hygiene is good", which is obviously true: but if you think that we weren't hygienic enough pre-pandemic, then say so! Don't try to wrap it up with anti-Covid measures, because it won't help much against Covid.
Read 7 tweets
Dec 17, 2021
Doing a thread about LFTs, PCRs and Bayes' theorem, so apologies in advance. But I feel like there's a lot of "if I get a positive LFT but then a negative PCR, do I have Covid or not?" going around, and people need to stop thinking about yes or no and think about risk levels.
Imagine that 1.5% of people have Covid (likely an underestimate). That's your prior probability.

You test 1,000,000 people. Given the prevalence, 15,000 of them ACTUALLY HAVE Covid. Your LFT is 99.9% specific and 70% sensitive (reasonable best guesses)
Of the 15,000 infected, it will correctly detect 10,500.
Of the 985,000 uninfected, it will incorrectly say 985 have Covid.

(It will also tell 4,500 infected people that they are not infected.)
Read 10 tweets
Oct 13, 2021
In which I say that, while there were many errors in the UK’s response to covid, the ur-mistake was gigantic overconfidence in uncertain science unherd.com/2021/10/the-me…
Via @graham8digits, a reminder that at least someone was making the same criticisms *at the time*. The UK response was based around incredibly precise manipulation of a chaotically uncertain reality
@Graham8digits this is something I was confused by. They were surprised by the Imperial 16 model showing that hospitals would be overwhelmed. But the flu plan expected 2.5% of infected to die, but *at most* 4% to go to hospital. So I think they expected most people to simply suffocate at home?
Read 4 tweets
Apr 27, 2021
in which I try to impart just how urgent it is that the West gets its spare vaccines to India (and starts making more vaccines to get ready for wherever the next India is) unherd.com/2021/04/india-…
Note: I use the IHME estimates for daily new cases in India in this piece. Someone's pointed out to me that those estimates are based on an implausibly low infection fatality rate, ≈0.05%, which would change the numbers: healthdata.org/sites/default/…
A plausible lower bound might be the Imperial model, which is around 1m new infections a day, rather than 10m ourworldindata.org/grapher/daily-…
Read 4 tweets
Sep 29, 2020
editors are great right up until they change the words you've written. if i'd wanted different words I'd have written them, god's sake
what if I *like* empty qualifiers, hmm, did you think about THAT before taking them all out
JUST BECAUSE IT MAKES THE PROSE BETTER IS NOT AN EXCUSE
Read 4 tweets
Sep 24, 2020
re Oxfam's carbon inequality report, which said the richest 1% accounts for 50% of emissions oxfamilibrary.openrepository.com/bitstream/hand… it says "We assume …that emissions rise in proportion to income". Doesn't that mean its findings are automatically implied? If the richest 1% get 50% of income…
…then the model will automatically say that they create 50% of emissions? I mean it's probably not *wrong*, it's just a bit weird, like saying "if we define the most handsome people as looking the most like Tom Chivers, then we find that Tom Chivers is the most handsome person"
which again is a slightly strange way of arriving at an obviously true conclusion
Read 5 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(