Remote

Aki Vehtari

@avehtari@bayes.club

mastodon 4.7.3

Professor in computational Bayesian modeling at Aalto University, Finland. Bayesian Data Analysis 3rd ed, Regression and Other stories, and Active Statistics co-author. #mcmc_stan and #arviz developer.

https://users.aalto.fi/~ave/

2551 Followers

249 Following

10 Posts

Joined November 03, 2022

Web page:

https://users.aalto.fi/~ave/

ORCID:

https://orcid.org/0000-0003-2164-9469

Bayesian Data Analysis 3rd ed:

https://www.stat.columbia.edu/~gelman/book/

Regression and Other Stories:

https://avehtari.github.io/ROS-Examples/

Aki Vehtari @avehtari@bayes.club

· 6mo ago

I'm looking for a post-doc to help organize Bayesian Data Analysis course https://avehtari.github.io/BDA_course_Aalto/Aalto2025.html (200 students) and to do research on Bayesian workflow https://users.aalto.fi/~ave/publications.html at Aalto University https://www.aalto.fi/en, Finland. Background in Bayesian topics needed. Up to five year contract possible.
Salary and occupational benefits are better than in academia in many other countries, and living costs are moderate. Finland is the world's happiest country 9th time in a row

avehtari.github.io

Bayesian Data Analysis course - Aalto 2025 – Bayesian Data Analysis course

Aki Vehtari @avehtari@bayes.club

· 31mo ago

Bob Carpenter told he had successfully used @mcmc_stan@bayes.club 's Pathfinder with "a big poli sci model with 1M latent parameters and 12M binary observations in 12 groups".

(For those who don't know Pathfinder yet, see https://jmlr.org/papers/v23/21-0889.html)

#Bayes

jmlr.org

Pathfinder: Parallel quasi-Newton variational inference

Aki Vehtari @avehtari@bayes.club

· 27mo ago

`posterior` package (Tools for Working with Posterior Distributions) has new CRAN release which includes additional Pareto khat diagnostics support and corresponding vignette illustrating the use for diagnosing reliability of estimating any expectation with empirical mean https://mc-stan.org/posterior/articles/pareto_diagnostics.html

See other enchantments in the NEWS https://cran.r-project.org/web/packages/posterior/news/news.html

#bayes #mcmc #rstats

mc-stan.org

Aki Vehtari @avehtari@bayes.club

· 47mo ago

Replying to

The main ideas are by @modrak_m@fediscience.org, who has been using SBC a lot in his own work, and everyone else did great work, too! The term in the previous literature is “simulation-based calibration”; here we have added the word “checking” to emphasize that these methods do not themselves produce calibration; rather, they measure departure from calibration. The code is available in SBC package https://hyunjimoon.github.io/SBC/ and experiments are at https://github.com/martinmodrak/sbc_test_quantities_paper #rstats

hyunjimoon.github.io

Simulation Based Calibration for Bayesian models

SBC helps perform Simulation Based Calibration on Bayesian models. SBC lets you check for bugs in your model code and/or algorithm that fits the model. SBC focuses on models built with Stan <https://mc-stan.org>, but can support other modelling languages as well.

Aki Vehtari @avehtari@bayes.club

· 25mo ago

Replying to

@rito@sigmoid.social Just tested, and it opens for me, so I don't know what to change

Aki Vehtari @avehtari@bayes.club

· 25mo ago

Replying to

The research would include collaboration, for example, with Andrew Gelman, @rlmcelreath@nerdculture.de, @paul_buerkner@fosstodon.org, and @mcmc_stan@bayes.club developer team and potential research visits with them

Aki Vehtari @avehtari@bayes.club

· 25mo ago

A bit more information about the forthcoming StanCon 2024 model selection tutorial in a Stan discourse post https://discourse.mc-stan.org/t/stancon-2024-model-selection-tutorial/36322

#Bayesian

discourse.mc-stan.org

Aki Vehtari @avehtari@bayes.club

· 31mo ago

Replying to

@sethaxen@bayes.club @mcmc_stan@bayes.club Pathfinder ran in realistic time with reasonable parameter estimates. Pareto khat is likely to be huge with 1M dimensions, even if the approximation would be close to true posterior (see Section 3.3 in PSIS paper). Bob is going to run MCMC, too, when he reported this had not yet done that.