Thoughts by Your Hommie
  • Home
  • Stats
  • Simulations
  • Books
  • All posts
  • Projects
  • About
Categories
All (3)
book-revisions (1)
prior-sims (1)
stats-cases (1)

All posts

Everything, newest first. Use the categories to narrow it down.

Separation makes logistic regression fail loudly in the warnings and quietly in the estimates

stats-cases

As a predictor gets closer to perfectly separating a binary outcome, the maximum likelihood estimate diverges. R warns about fitted probabilities, not about the coefficient — which is the number people actually report.

2026-08-12
5 min

A Normal(0, 10) prior on a logistic coefficient is not weak — it is a confident bet on certainty

prior-sims

Priors are chosen on the log-odds scale and interpreted as vague. Pushed through the inverse logit, a wide Normal prior puts most of its mass on probabilities of essentially 0 or essentially 1.

2026-08-11
6 min

The standard error, rewritten without the sampling-distribution detour

book-revisions

Most textbooks introduce the standard error by first constructing an imaginary population of repeated samples. That detour is where readers lose the thread, and it is avoidable.

2026-08-10
4 min
No matching items
Source Code
---
title: "All posts"
subtitle: "Everything, newest first. Use the categories to narrow it down."
listing:
  contents: posts
  sort: "date desc"
  type: default
  # Matches the ISO date on the post pages themselves. Listing `date-format` is
  # separate from the one in posts/_metadata.yml — setting only that one leaves
  # cards reading "Aug 12, 2026" next to posts reading "2026-08-12".
  date-format: "YYYY-MM-DD"
  categories: true
  sort-ui: [date, title]
  filter-ui: true
  # partial, not full. A full feed embeds each post's entire rendered body,
  # interactive-figure JSON included — that made blog.xml 7.2 MB.
  feed:
    type: partial
  fields: [date, title, description, categories, reading-time]
page-layout: full
title-block-banner: false
toc: false
---

© 2026 Batesthommie

Built with Quarto. Results are frozen at publication.