Tag: forecasting

All the articles with the tag "forecasting".

• A Hawkes Process Is a Poisson Process Whose Events Cause More Events

  11 Nov, 2026

  A Poisson process has no memory. A Hawkes process lets every event raise the rate of the next, and its ETAS version forecasts aftershocks. A worked example on the 2019 Ridgecrest earthquakes shows the likelihood, the trick that makes it fast in PyMC, and a clean-looking fit that was wrong.

• Election Night Is Bayes' Rule: What the Midterm Calls Did to a Frozen Forecast

  11 Nov, 2026

  I froze a forecast of the 2026 midterms before the polls closed, then updated it call by call. A call is an interval on a margin, and the calls that moved the whole map were the ones that said the polls were off.