TLDR: Honesty is the best policy, and don’t be a try-hard.
knite
Karma: 202
I understand that data collection is difficult and empathize with the people responsible for doing the work.
The thing is, SF used to publish everything as soon as they could! We accepted that numbers could be revised up or down as data was fully coded. This 5 day lag is IMO far on the wrong side of timeliness vs correctness.
Obvious next step: if there’s a lot of low hanging fruit like this, let’s find it? Have you considered using your LW/Twitter/blog to publicly solicit obvious, simple, and high leverage solutions to other big problems?
In dath ilan, it is virtuous to write more stories about dath ilan.
I expect agave to be generally preferred over table sugar and HFCS due to having a significantly lower glycemic index. I’m unfamiliar with Karo.
Something I’ve been wondering for a while: are organizations/journalists/individuals filing FOIA requests to get emails and other relevant documents about how the CDC and FDA made their COVID decisions?
Potentially interested!
Big picture, if your friend wants a different blend of upside-to-work, perhaps they should consider hiring someone to work 15-20 hrs/wk, freeing them up to do <5 hrs/wk of supervision?
This post is a bit hard to parse—please consider replacing “a.test” with something like “test.com/a” or “a.test.com/page” to clarify whether the issue is per-page caching or per-domain caching.
I posted my answer a bit late but this was a ton of fun!
Only four creatures have been known to do significant damage, implying that they’ve probably sunk some ships:
demon whales
nessie
merpeople
crabs
All attacks by these creature appear to be roughly the same relative frequency on a yearly basis.
Demon whales look the scariest, let’s max out at 20x oars: −20% for 20 gp.
Nessie has probably taken out a few ships but she rarely does more than 90%. Definitely worth a cannon: −10% for 10 gp.
Merpeople and crabs both have weird long tails. Let’s skip the merpeople entirely for 45 gp.
What’s sinking all these ships? Is it...murdercrabs? Let’s arm our carpenters, −50% for 20 gp.
We’ve never seen any of the other creatures come close to sinking a ship, so we won’t worry about sharks with lasers on their heads or mega-harpies. I have 5 gp remaining, which I keep as a fee for my services. Alternately, I’d consider trimming off 5 oars for a second cannon, but I’m not sure it’s worth it.
Two thoughts.
First, it wasn’t immediately clear that you meant within a range of [-1, 1], perhaps adding that to the graphic would help?
Second, this sounds like it generalizes as “trust your own opinion on any topic sigmoidally, scaling with your personal knowledge of it”—in other words, actively notice and reject your initial bias, until you have enough background to be truly informed, at which point you should trust your own judgment.
If you’re willing to write a data extraction script, and John Hopkins continues updating from a new source (that doesn’t otherwise publish raw data), you can find the numbers on the embedded sub-pages:
https://coronavirus.jhu.edu/embed/testing/state-data/testing_per_state_New_York.html
Regarding 14⁄15, I felt that we were probably under-reacting, but “general consensus” is tricky. We were in the home stretch of the Trump presidency so I figured the baseline odds of “consensus” on anything were extremely low.
I’m kicking myself on #16 - I don’t know enough about epidemiology to make such a strong guess.
Is the rule supposed to be symmetric around 50%? I used ln(p) - ln(.5) because Scott wrote:
“I scored these using a logarthmic scoring rule, adjusted so that guessing 50-50 always gave zero points.”
However, this doesn’t square with his second statement:
“Getting everything maximally right gives a score of about 14; guessing 50-50 for everything gives a score of 0, getting everything maximally wrong gives a score of negative infinity.”
Do you know what the correct scoring rule is?
Your explanation is fantastic!
Zeroing out from Zvi’s prediction, I am more optimistic and predict 4.3% national positivity rate.
I think that predictions are generally too pessimistic and that the rate of improvement will accelerate faster than the new strains matter and faster than the control systems kick in. I’ve been thinking this for a while and figure I should go on the record.
I agree! I would happily support these updates on Patreon, if that were an option. I understand that from Zvi’s perspective that could be problematic for any number of reasons, including possibly reducing motivation.
Zvi—since “money is not in any way the reason I write these posts”, how about favorite charities?
This seems worth a new post!
That being said, as I am an offensive person and wear that mantle publicly and without shame consider this your invitation to leave. You are responsible for what you choose to engage with. I am not responsible for your feelings and reactions. If you choose to stay then that’s on you.
Sorry bud, this is my post, so consider this your invitation to leave.
I did some initial exploration of the dataset and came to similar conclusions as others on the thread.
I then decided this was a good excuse to finally learn how to use LightGBM, one of the best-in-class tools for creating decision trees, and widely used in the data science industry. In other words, let’s make the computer do the fun part!
The goal was to output something like:
What I actually got:
I used default settings, transformed color/fangs/nostrils into 0-N categorical variables and marked them accordingly, then basically did “give me a regression with a single tree and 15 leaves”.
As others have mentioned, all gray turtles have fangs and weigh noticeably less (4-7 pounds), so this is obvious nonsense.
This tool is supposedly the non-AI state-of-the-art. It confidently fails with out-of-the-box settings. I remain baffled as to how anyone in tech ever gets anything done, myself included.