Shachar Hochman

New preprint with @rogierk.bsky.social @paulbuerkner.com - we introduce "relative measurement uncertainty" - a reliability estimation method that's applicable across a broad class of Bayesian measurement models (e.g., generative-, computational- and item response theory-models osf.io/h54k8

Exploring {ggplot2}’s Geoms and Stats – Stat’s What It’s All About

New blog post!

Ever wonder what geom_histogram is actually doing? How about geom_boxplot?

In celebration of the release of #ggplot2 4.0.0 (ggplot8?), I explore the relationships between the “geoms” and “stats” offered by the core {ggplot2} functions.

#rstats

Within-person factorial experiments, log(normal) reaction-time data | A. Solomon Kurz Causal inference with the GLMM, Part 1

New #rstats blog up!

solomonkurz.netlify.app/blog/2025-07...

This is the first in a new series discussing causal inference with experimental data using multilevel models. My basic case is g-computation is the way to go.

Dive in for code, visuals, and a clearer path through the log-odds fog → cogpsychreserve.netlify.app/posts/logist...

#NLP #Kaggle #marginaleffects #BayesianStatistics #DataScience #SignificantTesting

2/3

• NLP + PCA to capture toxicity/incoherence
• Cohen’s d ➡️ log-odds priors in one line using #brms
• #marginaleffects → 0–100 % probability shifts you can explain
• Inference with HDI-ROPE. It flags which effects are big enough to matter. Great for researchers and anyone shipping spam filters!

1/3 New post up! 📝 I took the workhorse 🔧 of binary modeling—logistic regression—and gave it a Bayesian tune-up using a Kaggle SMS-spam dataset.

Thanks Laura! 🙏 I analyzed vertical-face tasks (6 variants across SOAs) from subjects with mouse responses only. The Preprocessing details are in the post’s collapsible section 😊. Grateful for your work—DM anytime!

Thank you! 😊 While latent correlations are possible in Stan via custom likelihoods (modeling latent Gaussian variables), it's quite involved. For 95% of cases, I recommend the simpler brms approach: model questionnaires as predictors of task effects using condition-by-questionnaire interactions.

6/6 Thanks to @solomonkurz.bsky.social for statistical inspiration, @natehaines.bsky.social for works that influenced my approach, and @almogsi.bsky.social & @mattansb.bsky.social or thoughtful feedback!

#BayesianStatistics #ReliabilityAnalysis #CognitiveScience

5/6 The implications go beyond this single task. Many measures in psychology (and beyond) might be more reliable than we thought—we need to preserve and properly model the information in trial-level data.

4/6 This visualization shows the transformation when the same data is analyzed with trial-level Bayesian methods instead of traditional aggregation:

3/6 I implemented two Bayesian approaches in #brms:

@jeffrouder.bsky.social & @juliaha.bsky.social's variance decomposition
@gangchen6.bsky.social's approach

Both show substantially higher reliability than traditional analyses.

2/6 Recent research by @irenexu.bsky.social claimed the emotional dot-probe task lacks reliability for individual differences research. I wanted to see if more sophisticated analysis methods could tell a different story.

1/6 Hello Bluesky! 👋 Excited to join this community and share my new blog. First post: Using Bayesian hierarchical models to rescue "unreliable" cognitive tasks, with the dot-probe task as my case study. cogpsychreserve.netlify.app/posts/dotpro...