As in bio, have now mostly given up fighting through this spammy, bot-filled mess. May return if improved (e.g., foreclosure, change of ownership) but doubt it.

RT @ScarredForLife2: Feast your eyes on this gorgeous 1979 animation by Roger Mainwood (Heavy Metal, Stressed Eric, Ethel & Ernest), commis…

@salonium Is that COVID nudging Hong Kong out of the top spot for a couple of years to 2022?

@tnewtondunn More context: the US spends 2.9%.

@EdConwaySky A further difference is that the solar chart starts at zero. You need to heavily truncate the y-axis of the evil twin in order to make it look sufficiently evil.

@whippletom But what if this apocalypse happens before 7pm, are we to die not even knowing why?

RT @RichardHanania: If you haven't looked into their claims, you are always going to underestimate just how much and how blatantly anti-vax…

@Saraht0n1n I find the paper and notation very hard to follow but doesn’t it boil down to showing that classes of deterministic hidden variable theories are possible? Which we’ve known since de Broglie-Bohm? Still got to be non-local to reproduce QM, which we’ve known since Bell.

I too would like to register my concerns about the dangers of creating an elitist, comfortable echo-chamber, and so will be posting exclusively on MILF Messenger from now on.

@HarryDCrane @Polymarket Actuarial tables for a 78.4 year old male account for a bit more than a third of that (1 year prob of death = 0.055, so three month probability ~0.014). The rest ... ?

@TheStalwart You need the whole distribution. E.g., this summarises as a 50:50 chance, but is giving a high probability to a landslide one way or the other. Once you have the distribution, you can look at the actual results and use Bayes' theorem to update your trust levels in each model.

@ChapelGuilt Well, yes. That's why I wrote a whole section about it and reversed the conclusion as a result.

US Election: Humans are very bad at thinking about bimodal probability distributions, especially with lots of correlation amongst the parts. Twelve days out, I think that's exactly what we've got. Complete with squeaky-bum predictions.

Storms of criticism aimed at pollsters as the US election looms. Four that actually make sense: 1) Track record with Trump. 2) Flooding the zone with bad polls. 3) The ground game. 4) Weighting by recalled vote. Are they real, and do they matter?

@fchollet This seems to assume that the universe simulator is running on classical, not quantum, hardware.

How to decide which US polling model is best - they all basically say it's 53:47 right now? One approach: Which one makes you the most money on the betting markets? ($/free preview).

@markjrieke Dynamic Linear Modellers unite.

What would US election polling models look like if you stripped out some of the (frankly, daft) adjustments - particularly, a candidate’s “convention bounce”? ($, free preview)

@whippletom @d_spiegel ... with the monetisation achieved by selling all my data to aggressive telemarketers.