drkowal.github.io/lmabc/

Facilitating heterogeneous effect estimation via statistically efficient categorical modifiers Categorical covariates such as race, sex, or group are ubiquitous in regression analysis. While main-only (or ANCOVA) linear models are predominant, linear models that include categorical-continuou...

Paper and code/documentation below:

doi.org/10.1080/0162...

...doing so can introduce biases and make interpretations more difficult. The surprise is that a method designed to fix those problems actually gives some fantastic statistical efficiency results.

The short version: you can estimate heterogeneous effects without compromising "overall" effects.

Statistics that tell the whole truth? It’s as easy as ABC | Cornell Chronicle A Cornell statistics expert has come up with a method he believes can boost statistical power and significantly reduce bias – vital for research involving outcomes that differ by socioeconomics, race,...

Thanks for the story @cornellcals.bsky.social! This work is about one of the simplest, most fundamental, and most widely-used statistical methods: OLS linear regression. We often use OLS linear regression to estimate heterogeneous (or subpopulation) effects. But...

news.cornell.edu/stories/2026...

Otherwise a main difference comes from the smoothing parameter. With full Bayes, it's just a prior precision that's easy to learn jointly with all other parameters. But penalized estimates would fix this using (G)CV or REML and then need other (asymptotic) arguments to get UQ.

Penalties usually motivative priors but you can view it the other way too. If you have a Bayesian model (prior + likelihood) but only take the posterior mode, then the prior is exactly a penalty. Or you do full Bayesian inference and get UQ but then you've made explicit distributional assumptions.

Jan 27 COPSS-NISS Webinar, The Role of Bayesian Statistics in an Age of AI, will feature David Dunson (Duke U.) and Xuming He (Washington U. in St. Louis) exploring Bayesian methods for transparency, interpretability, and decision-making in modern AI systems.

Details 👇
www.niss.org/sites/defaul...

link 📈🤖
Efficient Bayesian inference for two-stage models in environmental epidemiology (Larin, Kowal) Statistical models often require inputs that are not completely known. This can occur when inputs are measured with error, indirectly, or when they are predicted using another model. In environm

Dear ISBA members,

It is with great pleasure that we announce the completion of this year’s round of the Blackwell–Rosenbluth Award by j-ISBA, and that I have the honor of presenting our 2025 winners:

Uncovering Dynamic Relationships Between SARS-CoV-2 Wastewater Concentrations and Community Infections via Bayesian Spatial Functional Concurrent Regression

@tandfresearch.bsky.social #COVID #DataScience www.tandfonline.com/doi/full/10....

Monte Carlo Inference for Semiparametric Bayesian Regression Data transformations are essential for broad applicability of parametric regression models. However, for Bayesian analysis, joint inference of the transformation and model parameters typically invo...

The update includes:
- faster computing
- fewer dependencies
- new functionality for variable selection in semiparametric regression
- better handling of the transformation (now properly location-scale identified)

The paper is also open access: www.tandfonline.com/doi/full/10....

SeBR: Semiparametric Bayesian Regression Analysis Monte Carlo sampling algorithms for semiparametric Bayesian regression analysis. These models feature a nonparametric (unknown) transformation of the data paired with widely-used regression models inc...

New update to the SeBR package! Efficient, fully Bayesian inference for transformed linear/quantile/GP regressions. The learned transformation lets these models apply far more broadly:
- Data that are skewed, heavy-tailed, multimodal...
- Continuous real, positive, or [0,1] data

Warped Dynamic Linear Models for Time Series of Counts Dynamic Linear Models (DLMs) are commonly employed for time series analysis due to their versatile structure, simple recursive updating, ability to handle missing data, and probabilistic forecasting. ...

doi.org/10.1214/23-B...

Bayesian Analysis journal "most read articles"

Join the crowd of folks reading Brian's paper on count time series models!

A dynamic copula model for probabilistic forecasting of non-Gaussian multivariate time series Multivariate time series (MTS) data often include a heterogeneous mix of non-Gaussian distributional features (asymmetry, multimodality, heavy tails) and data types (continuous and discrete variables)...

New paper on probabilistic forecasting for multivariate time series! Work led by my former PhD student, John Zito:

Immensely sad that Herman has passed away.
He took me into the Bayesian community at the Valencia meeting in 1998. Very encouraging discussant of my thesis. Always helping the younger and willing to share his time and knowledge.
He will be missed by many.

Obituary:
esobe.org/Herman_van_D...

Since this work has gotten some attention, reposting for all the new folks here!

www.nationaltribune.com.au/new-statisti...

Please add, thanks!

We're hiring! Positions available at the Asst/Assoc level. Wonderful collaborative opportunities with @cornellcals.bsky.social and throughout the university. And we're moving to a brand new building for data/computing/info sciences this summer! Please apply:

academicjobsonline.org/ajo/jobs/28628

Please add me, thanks!

Please add me, thanks!

I've never had anything "trending" before...thanks
@cornellcals.bsky.social and Phys.org for getting the word out!
#Statistics #AcademicChatter #academicsky #Research #Science

Please add me, thanks!

I also built a website with examples and software documentation (R package 'SeBR' is on CRAN):

#rstats

drkowal.github.io/SeBR/

Monte Carlo Inference for Semiparametric Bayesian Regression Data transformations are essential for broad applicability of parametric regression models. However, for Bayesian analysis, joint inference of the transformation and model parameters typically invo...

Since I'm new here, I'll link to my most recent paper (open access!). It's a blend of Bayesian regression, data transformations, and computing.

#bayes #stats

doi.org/10.1080/0162...