Awesome

Learning Bayes

Just one of the things I'm learning. https://github.com/hchiam/learning

intuition and explanation:

How to use Bayes's rule to update beliefs when given new evidence (hint: multiply ratios): https://brilliant.org/courses/knowledge-and-uncertainty/bayesian-thinking/bayes-rule/1/ and https://brilliant.org/courses/knowledge-and-uncertainty/bayesian-thinking/cause-bayes-rule/1/

• p-value = p(results|H0)

  • small p-value <-- large effect OR large sample size!
  • instead of using "p-value < 0.05 is good", use Bayes' rule to weigh both evidence for (H1) and against (H0) - accounting for baseline rates, and multiple possible causes for an observation (don't forget the possibility of independent concurrent causes), i.e. observations given actuals

• google "causal Bayesian network"

backups of some of my old diagrams:

A indep B ⇒ p(A & B) = p(A) * p(B):

A not indep B ⇒ p(A & B) = p(A if B) * p(B):

A xor B ⇒ p(A or B) = p(A) + p(B):

A not xor B ⇒ p(A or B) = p(A) + p(B) - p(A: & B)