Stan

@mcmc_stan

Making Bayes fast since 2011

Vrijeme pridruživanja: veljača 2013.

Tweetovi

Blokirali ste korisnika/cu @mcmc_stan

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

  1. Prikvačeni tweet
    14. stu 2019.

    RStan should now work on Catalina with new version of 's toolchain installer for R . Thanks to everyone who helped troubleshoot!

    Poništi
  2. 2. velj
    Poništi
  3. proslijedio/la je Tweet
    26. sij

    Our preprint is out! 😀 We modelled the evolution of eczema severity and how patients respond to treatments from a Bayesian and mechanistic perspective, using Stan

    Poništi
  4. proslijedio/la je Tweet
    27. sij

    Computing a test & roll sample size requires information about the range of average response you expect from your treatments. Using data on the past performance of similar treatments, you can estimate priors with an HB model using tools like and . 14/17

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    27. sij

    Check out this preprint on hierarchical bayesian signal detection models by : They also have an accompanying R package that implements the models in language.

    Poništi
  6. proslijedio/la je Tweet
    27. pro 2019.

    New loo CRAN release has LOO with subsampling and approximations to posterior distributions which make LOO much faster for large data. See a vignette for examples and references.

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    13. sij

    Over the xmas and new year's break, I wanted to learn about state space models. Forked 's repo where he tried to recreate the models from using . Super helpful! My modifications & extensions into Ch8:

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

    Running models. Only doing 1000 warmup and an additional 1000 iterations and thinning in half on 6 chains for N-mixture model. No divergences and good tail ESS. So weird after running 100,000 in JAGS previously.

    Poništi
  9. proslijedio/la je Tweet
    12. pro 2019.

    This past semester I taught a graduate-level course on Bayesian inference and applied statistical modeling. I'm making all of the computer labs publically available here: Let me know if you need help with solutions! Rmarkdown RStanARM.

    Poništi
  10. proslijedio/la je Tweet
    Odgovor korisniku/ci

    [pt3] if you cannot get a very reliable measure, you should take the (un)reliability into account. Bayesian models do this nicely. linked our pub using also Mplus has facilities and this pub has power sims:

    Poništi
  11. proslijedio/la je Tweet
    17. pro 2019.
    Poništi
  12. proslijedio/la je Tweet
    19. pro 2019.

    From July 2015 to this week, RStan for has been downloaded from RStudio CRAN mirror one million times (loo 630K times). Python users love much more downloading as PyStan gets 700K+ downloads just in the last month.

    Prikaži ovu nit
    Poništi
  13. proslijedio/la je Tweet

    Stan ecology community, it's happening! Very much looking forward to putting this together with all of the wonderful people in the community already using Stan!

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet

    If you see something weird about open source software, say something. found a weird discrepancy in some results, posted about it on discourse with a simple reproducible example, and the bug was found super fast!

    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    13. pro 2019.

    A little late announcing this. In early Nov I became one of the members of the Stan Governing Body & I'm very excited about contributing to this amazing open-source project in this role.

    Prikaži ovu nit
    Poništi
  16. proslijedio/la je Tweet
    10. pro 2019.

    Bayesian longitudinal models using additive GPs including uncertainty in disease effect times and heterogeneous effects by , Henrik Mannerström, I and . lgpr package built on top of at Preprint at

    Poništi
  17. proslijedio/la je Tweet
    9. pro 2019.
    Poništi
  18. proslijedio/la je Tweet
    6. pro 2019.

    An example of exploring joint posterior space for some normally distributed data using in . This represents only 150 samples/chain (for the sake of brevity), which means most chains have found the target after only about 50 iterations. So cool.

    Poništi
  19. proslijedio/la je Tweet
    9. pro 2019.

    Interested in ? We will run the 2nd edition of the course" An introduction to computational Bayesian methods" with and at the next March. If interested, please see:

    Poništi
  20. 26. stu 2019.

    Bob Carpenter's nice list of recent progress in the Stan community

    Poništi
  21. proslijedio/la je Tweet
    25. stu 2019.

    New coding example on my website: "NFL field goal attempts: modeling the long shots" In which I take a *three*-parameter model to 50,741 coin flips. H/T to for suggesting the informative prior.

    Prikaži ovu nit
    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:

    ·