Multiple Hypotheses and Total Probability
What You'll Learn
You'll learn how to use Bayes' Theorem to compare multiple competing explanations for the same piece of evidence, determining which hypothesis is most likely.
The Core Idea
So far, you've used Bayes' Theorem to update your belief in a single hypothesis (like "patient has disease" vs "patient doesn't have disease"). But real-world problems often involve choosing between several competing explanations.
Example: An email arrives. It could be spam, a personal message, or a work-related message — three hypotheses. The evidence (words like "meeting" and "urgent") should help you figure out which is most probable.
Extending Bayes to Multiple Hypotheses
When you have hypotheses H₁, H₂, H₃, etc., and observe evidence E:
-
Calculate the posterior for each hypothesis using Bayes' Theorem:
- P(H₁|E) ∝ P(E|H₁) × P(H₁)
- P(H₂|E) ∝ P(E|H₂) × P(H₂)
- P(H₃|E) ∝ P(E|H₃) × P(H₃)
-
Use the Law of Total Probability to normalize: The denominator P(E) equals P(E|H₁)×P(H₁) + P(E|H₂)×P(H₂) + P(E|H₃)×P(H₃) + ...
-
Compare posteriors to see which hypothesis is most supported by the evidence.
Why This Matters
This approach lets you rank multiple explanations instead of just accepting or rejecting one. It's the foundation of classification algorithms and diagnostic systems that choose between many categories.
Key Takeaway: When multiple hypotheses compete to explain the same evidence, calculate each hypothesis's posterior probability using Bayes' Theorem with the Law of Total Probability for normalization, then compare to find the most likely explanation.