Profile avatar
jordannafa.bsky.social
Bayesian Statistician and Data Scientist in the Gaming/Entertainment Industry | Bayesian Statistics, Causal Inference, R, Python, Stan, Decision Theory, Guitar
937 posts 1,057 followers 240 following
Regular Contributor
Active Commenter
Posts 20 Comments 30

Population-level linear regression model is kind of terrible. But add a few random effects, a couple interactions, and some relatively informative priors to prevent things from blowing up and you end up with something that has surprisingly decent performance as an out of sample prediction tool.

submitted 1 day ago • 0 comments

You too can basically double your out of sample prediction performance with this weird trick we call *checks notes* cleaning your data before passing it to your model...

submitted 1 day ago • 0 comments

Going to Costco today to buy six months worth of toilet paper because this gives March 2020 vibes

submitted 4 days ago • 0 comments

50% of men and 40% of women are consistently trash.

submitted 4 days ago • 0 comments

Figured out how to invoke Stan from basically any programming language in existence today, so that was cool. Basically makes the language for writing models fully independent from whatever the rest of the system is.

submitted 6 days ago • 3 comments

priorsense 1.1.1, for prior and likelihood sensitivity analysis, is now in CRAN (n-kall.github.io/priorsense/) There is now an easier way to select which priors are power-scaled using the new prior tags feature in brms (2.22.11+), which allows focusing the analysis on specific priors. 1/3

submitted 7 days ago • 1 comment

Motivating Bayesian inference from a frequentist perspective is a fun exercise. If your goal is a reliable estimator with good frequency properties the provides accurate and conservative estimates, HMC with a moderately regularizing prior is a remarkably efficient means to achieving that end.

submitted 7 days ago • 1 comment

The fun part of this is that the actual logic is pretty much the same in both places

submitted 7 days ago • 0 comments

"Everyone should just become a Bayesian" is always my favorite answer

submitted 8 days ago • 1 comment

Yeah, this accurate

submitted 9 days ago • 1 comment

😮‍💨

submitted 9 days ago • 0 comments

Imagine you have a Bayesian time series-esque model where time is modeled via smoothing splines. What is the proper approach to cross validation in such a case given splines are basically random effects? #statistics

submitted 9 days ago • 4 comments

This is what happens when you elect Republicans. Every. Single. Time.

submitted 10 days ago • 0 comments

Everytime someone says the confederacy was "their heritage," what they're really saying is that their ancestors were traitors and the single greatest mistake this country ever made was choosing not to publicly hanging every last one of them.

submitted 10 days ago • 1 comment

Is this a new record for shortest tenure as SecDef?

submitted 10 days ago • 0 comments

>The earliest statisticians would spend hours thinking about their assumptions and models because the computations themselves might have taken days. Today we call those people Bayesians

submitted 11 days ago • 3 comments

In part 2 of the series on poll weighting leading up to AAPOR's conference, I recreate the results of Little & Vartivarian's 2005 paper, Does Weighting for Nonresponse Increase the Variance of Survey Means? www.thedatadiary.net/posts/2025-0...

submitted 11 days ago • 0 comments

Short term, you shoot yourself in the foot. Long term, you develop gangrene and die of fucking sepsis.

submitted 12 days ago • 0 comments

It annoys me that there isn't a {ggdist} equivalent in python, and I'm too lazy to try and implement one, so still using R for all visualization

submitted 13 days ago • 1 comment

I can't get past how fucking stupid someone has to be to simultaneously believe that "AI" is going to automate 50% or more of human labor in the next 20 years, but also that the birth rate needs to increase dramatically and yet, there are people who believe these things simultaneously 🤦‍♂️

submitted 13 days ago • 0 comments