Open Tokazama opened 5 years ago
Please tell me more about how this would work, and what properties it would have
I was thinking something like
struct DistributionTerm <: AbstractTerm
term::AbstractTerm
dist::Distribution
end
Where formula syntax may be @formula(y ~ x, x ~ Normal())
.
Models that utilize posterior probabilities could use the entire term and if a user mistakenly puts it in a linear model it would still have the default term information.
There are a lot of smaller details to consider, but at at this point there's an explosion of probabilistic programming going on. It would be nice if they could start to rely on a common formula syntax.
Are there any plans to create a
DistributionTerm
and if so is this the place to get that rolling or would that be a separate package likeDistributionModels.jl
?