In lectures/prob_meaning.md, the lecture updates the posterior sequentially (one Bernoulli observation at a time), but the exercise solutions use the binomial likelihood over n draws. The equivalence between these two approaches — that updating n times after each Bernoulli observation gives the same posterior as updating once using a Binomial over n draws — is never stated explicitly.
This follows from Beta-Binomial conjugacy: the posterior depends only on the sufficient statistic (k heads in n flips), not on the order of updating. A brief remark after the n-step posterior derivation would make this connection clear for the reader.
In
lectures/prob_meaning.md, the lecture updates the posterior sequentially (one Bernoulli observation at a time), but the exercise solutions use the binomial likelihood over n draws. The equivalence between these two approaches — that updating n times after each Bernoulli observation gives the same posterior as updating once using a Binomial over n draws — is never stated explicitly.This follows from Beta-Binomial conjugacy: the posterior depends only on the sufficient statistic (k heads in n flips), not on the order of updating. A brief remark after the n-step posterior derivation would make this connection clear for the reader.