Alex

YouTube video by PyData Scaling Probabilistic Models with Variational Inference

Here is the recording of my talk

PyData Berlin 2025: Introduction to Stochastic Variational Inference with NumPyro

Notebook: juanitorduz.github.io/intro_svi/

youtu.be/wG0no-mUMf0?...

#pydata #berlin #bayes

#rstats Please welcome Jarl, a new R linter.

Jarl is a CLI tool with extensions in VS Code, Positron, and Zed. It can check thousands of lines of R code in milliseconds.

Jarl provides several output formats, a Github Actions workflow, and more.

Blog post: www.etiennebacher.com/posts/2025-1...

Quickly added support for @apoorvalal.com 's duckreg - took approximately 60 minutes!

github.com/py-econometr...

Admittedly, this was 99% Dirk's work, so all praise should go to him =)

Btw, also works with Stata! py-econometrics.github.io/maketables/d...

@gaborbekes.bsky.social @arthurturrell.bsky.social @vincentab.bsky.social

Our py-econometrics org has a new package: `maketables` is now installable from PyPi! You can create regression tables for PyFixest, statsmodels, linearmodels, and save them as html, latex, or word.

github: github.com/py-econometr...
docs: py-econometrics.github.io/maketables/

Looking forward to the follow-up post on fixed effects in GLMs =)

Trying to learn more about fixed effects. I wrote this for me, maybe this is useful for you too dpananos.github.io/posts/2025-1...

Sorry for all the linkedin post, they just transfer you to the docs / github (I posted on lk first).

- You can find the full changelog here: lnkd.in/eEdzuVvq
- PyFixest on github: lnkd.in/eJRtm_rz
- maketables on github: lnkd.in/eYhAejYh

It currently supports pf and sm. Please check it out and contribute by adding support for your favorite Python estimation classes! Some that we think would be of interest: linearmodels, lifelines, Bambi, rdrobust, DoubleML, meta learners. You can find the package here: lnkd.in/eYhAejYh

We believe that there is a lot of value to unlock by providing a "𝗺𝗲𝘁𝗮 𝗿𝗲𝗴𝗿𝗲𝘀𝘀𝗶𝗼𝗻 𝘁𝗮𝗯𝗹𝗲" package for the Python scientific eco-system similar to what R has in 𝗺𝗼𝗱𝗲𝗹𝘀𝘂𝗺𝗺𝗮𝗿𝘆. So we have spun out PyFixest's etable functionalities into a standalone package, 𝗺𝗮𝗸𝗲𝘁𝗮𝗯𝗹𝗲𝘀.

This strategy is quite amenable to be run on the GPU - for "hard" problems where the alternating projections algorithm struggles to converge, running the demeaning on a 𝗚𝗣𝗨 can provide orders of magnitude of speedups.

We've added 𝗮𝗻𝗼𝘁𝗵𝗲𝗿 𝗯𝗮𝗰𝗸 end for 𝗱𝗲𝗺𝗲𝗮𝗻𝗶𝗻𝗴 𝗳𝗶𝘅𝗲𝗱 𝗲𝗳𝗳𝗲𝗰𝘁𝘀: instead of demeaning via the alternating projections algorithm, the new back end implements the Frisch-Waugh-Lovell theorem directly on 𝘀𝗽𝗮𝗿𝘀𝗲 𝗺𝗮𝘁𝗿𝗶𝗰𝗲𝘀.

We now support time series, panel 𝗛𝗔𝗖 and DK 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱 𝗲𝗿𝗿𝗼𝗿𝘀 (which was one of the most requested features, so I am very happy we have it now). It was also a very challenging PR, with very-tricky to debug numpy multi threading floating point errors.

First, a couple of weeks ago, the 𝗳𝗶𝘅𝗲𝘀𝘁 dev team released 𝟬.𝟭𝟯, which comes with a 𝗳𝗲𝘄 𝘀𝗺𝗮𝗹𝗹𝗲𝗿 𝗯𝗿𝗲𝗮𝗸𝗶𝗻𝗴 𝗰𝗵𝗮𝗻𝗴𝗲𝘀. We have decided to 𝗳𝗼𝗹𝗹𝗼𝘄 𝘀𝘂𝗶𝘁 and adjust PyFixest accordingly. The most important change is that by default, all standard errors are "iid".

I am very happy that 𝗣𝘆𝗙𝗶𝘅𝗲𝘀𝘁 𝟬.𝟰𝟬.𝟬/𝟬.𝟰𝟬.𝟭 is now available on PyPi! The release includes a few technical changes and exciting new features =)

Oh interesting, this does not look good. For me everything has been working well, I just upgraded to the newest fixest release via conda-forge a few weeks ago without any problems (as it happens, on windows!).

No global R or Python installation needed! And if you have these, no interference with your global installations.

Big fan. pixi.sh/latest/

Got a new PC and getting back to pyfixest dev is a breeze with pixi (via conda-forge, which seems to get too little love?). This just works: clone pf, install pixi, type pixi r tests in the shell, and then pixi installs Python, R, all Python and R deps, all R deps, and then starts running tests.

fixest: Fast Fixed-Effects Estimations Fast and user-friendly estimation of econometric models with multiple fixed-effects. Includes ordinary least squares (OLS), generalized linear models (GLM) and the negative binomial. The core of the p...

#rstats #econsky

fixest v0.13.0 is finally out!

It's still about making OLS and GLM estimations easy.

Some major changes:
- *default* VCOV becomes iid always!
- singletons are removed by default!

See all the changes here:
github.com/lrberge/fixe...

From the fixest NEWS file: the new default VCOV is iid for all estimations. To change the default to the way it was, place setFixest_vcov(all = "cluster", no_FE = "iid") in your .Rprofile.

#rstats #econsky PSA: The next release of `fixest` will include some important changes (plus cool new features).

E.g. Fixed-effects regs will now default to 'iid' SEs rather than clustered. github.com/lrberge/fixe...

You can install and test drive the dev version from R-universe; see the README.

Next step is to rewrite the vignette (more to the point & I need to make a better case for the methods usefulness in business contexts).

Also works with "negative" effects:

Vis method for decomposition now merged to main, feedback welcome!

Love to see this, of course =)

In contrast to uv, it is cross-language and also allows to pin down non-python dependencies in the same env (R, Julia, etc). And it also has nice "pixi run command" options (you can even define tasks as in justfiles, which I personally use a lot). pixi.sh/latest/

For pyfixest's development, we've been using pixi for a while, which I really enjoy. More or less it is the conda-forge (free & open source) equivalent to uv (it actually runs uv for PyPi dependency resolution). Really fast both for PyPi & conda-forge dependency resolution.

This I will still have to explore (as matplotlib does not always auto-scale). Generally there are two options to handle this as of now - users can keep / drop covariates via a function arg, or simply combine many of them into a single "combined" covariate in the estimation process.