x
Likelihood ratio — LessWrong