search

stephenslab / ebnm

R package to fit Empirical Bayes Normal Means model.
https://stephenslab.github.io/ebnm/
12 stars 9 forks source link

add point_cauchy prior family #51

Open willwerscheid opened 3 years ago

willwerscheid commented 3 years ago

interface to EbayesThresh

stephens999 commented 3 years ago

does EbayesThresh do point-cauchy? I thought it was either point-laplace or cauchy? maybe i'm misrembering

willwerscheid commented 3 years ago

yes, it's point-Cauchy (or rather, a mixture of a pointmass and a scale mixture of normals with polynomial tails that the authors call "quasi-Cauchy")

pcarbo commented 3 years ago

There are two options: prior = "cauchy" and prior = "laplace". Only the laplace prior is allowed for case of heterogeneous variances.

willwerscheid commented 3 years ago

right, but the "laplace" option is actually point-laplace and "cauchy" is as i've described above

pcarbo commented 3 years ago

Yes, you are right to clarify that point.

stephens999 commented 3 years ago

thanks for clarification - yes, i was wrong

willwerscheid commented 3 years ago

EBayesThresh doesn't return a posterior second moment or sampler, so adding the point-Cauchy family might be more trouble than it's worth