I’m a bit confused here, by “the original prior” you mean the “estimated prior” with beta…

Original article was published on Artificial Intelligence on Medium


I’m a bit confused here, by “the original prior” you mean the “estimated prior” with beta distribution, right? Not the true data generating clap_prob of 0.3? I thought the purpose of MAP is trying to get a good estimate of the true prior. If we keep the data generating clap_prob fixed at 0.3, then the more data we gather, the peak of the posterior on theta will move closer towards 0.3? As will the peak of the estimated prior using beta distribution if we change alpha and beta accordingly.