Content-type: text/html Downes.ca ~ Stephen's Web ~ How to think like a Bayesian

Stephen Downes

Knowledge, Learning, Community

This article is really wordy but it also addresses a really important concept: prior probabilities, and in particular, Bayes's theorem. Scroll down for the 'key points'. There are some good examples; I like this one: suppose there's a disease that affects 1 in a hundred people. But there's a test for the disease, and it's 90% accurate. So you test someone, and it indicates that the person has the disease. How likely is it that they actually do have the disease? You might be tempted to say '90%'. But what Bayes's theorem tells us is that because the prior probability is so low - only one in a hundred - that for every one person who actually has the disease there are 10 people who were mislabeled by the test. So even if the test is accurate there's only about an 11 percent chance that they actually have the disease. Here's an online Bayes Calculator you can play with.

Today: 0 Total: 11 [Direct link] [Share]


Stephen Downes Stephen Downes, Casselman, Canada
stephen@downes.ca

Copyright 2024
Last Updated: May 30, 2024 02:00 a.m.

Canadian Flag Creative Commons License.

Force:yes