Charles R. Twardy

@ctwardy

Applied epistemology, Bayesian weirdness, bad puns.

Virginia
Vrijeme pridruživanja: rujan 2010.

Tweetovi

Blokirali ste korisnika/cu @ctwardy

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @ctwardy

  1. prije 21 sat

    And let us know what you think of the new survey format!

    Poništi
  2. 31. sij

    Hey - do any forecasters use GRIM, Statcheck, etc.? Test the digit distribution? WebPlotDigitizer? Other tools similar to those mentioned here?

    Poništi
  3. 31. sij

    But (b) implies c) smaller average payout per market. Revising budget now to minimize the impact - new rules posted next week.

    Prikaži ovu nit
    Poništi
  4. 31. sij

    Starting in Round 6, we will follow the larger SCORE project to a unified definition of "replication". Main effects: a) Surveys get shorter, b) More markets resolve - from R6, not just "direct" replications count!

    Prikaži ovu nit
    Poništi
  5. 31. sij

    This is good news for forecasters who use the scite widget.

    Poništi
  6. 30. sij

    An elegant twin-book review essay by Philosopher Philip Kitcher on replication, fraud, and slow science [journalism]: Has Science Journalism Helped Unmask a “Replication Crisis” in Biomedicine?

    Poništi
  7. 30. sij
    Poništi
  8. proslijedio/la je Tweet
    28. sij
    Poništi
  9. 28. sij

    It was a problematic survey - "opposite" answers A & F were both true. S makes sound critiques, and I wonder a bit about H&I carefulness. But most interesting is H&I's response to his Bayes frame: they simply think his skeptical prior not enough so. A defensible disagreement.

    Poništi
  10. 26. sij

    Nearly 100 participants in Round5. And nearly 2,500 trades. If you haven't looked at Round 5 yet, jump in. As always, "Edit out of outrage" as says - fix what's most wrong. A good start: review the least-traded claims. (Default sort.) They look OK? Look into 1?

    Poništi
  11. proslijedio/la je Tweet
    24. sij

    Yes, we have over 1000 people registered for our project. And yes, it is a small group of active participants ... less than 10% of you are forecasting in Round 5 markets. If you haven't made your first trade yet, why not start today?

    Poništi
  12. 23. sij
    Prikaži ovu nit
    Poništi
  13. 22. sij

    2019 vs 1997: "randomization ...up from 10-30%, power calculation from 2-10%, reporting sex from 22-37%, and antibodies ...43% [vs] 12%. So, ...better, but ...half of papers don’t tell you the sex of experimental subjects and how they are divided into groups. "

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    21. sij

    Quiet day in the Round 5 markets...a great day to jump in and make a trade! Did you know that all the claims you reviewed in the surveys are available in the markets? Find them under All Claims and show My Surveys.

    Poništi
  15. 21. sij

    Nifty. Adding decorator makes Python competitive with C++ on the recursive N-queens problem. "Develop in ordinary Python, identify computational bottlenecks, and use Numba to remove them."

    Poništi
  16. 21. sij
    Poništi
  17. 21. sij

    "People who do well..." - refers to published work on Bayesian Truth Serum: people who answer "a" but correctly predict most people will say "b" probably know something. We won't *know* if that's true in our data until mid-2020. But it's encouraging.

    Prikaži ovu nit
    Poništi
  18. 21. sij

    In our surveys, two questions ask how other people will respond - these can be evaluated immediately. People who do well on these meta-Qs tend to do well on the main Qs. Our peer-prediction AUC score on main Qs correlates 70-86% with AUC on these meta Qs.

    Prikaži ovu nit
    Poništi
  19. 18. sij

    Your surveys only affect the survey average - it has no voice in the market unless you speak for it. Re-use your work! Nudge the market towards the truth. Just once, or better, a little each day. Find and fix the claims that are crazy wrong. Share your knowledge!

    Poništi
  20. proslijedio/la je Tweet
    18. sij

    Lots of new traders in the Round 5 markets since yesterday! If you haven't already...join in! You've got free points to invest, and you'll help build our dataset...for science!

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·