compatibility with Turing.jl?

[fkrauer]

Hi

That's an interesting package. Does RunDIME accept a Turingmodel for the LogProb?

Cheers, Fabienne

[gboehl]

Hi Fabienne,

Thank you for your interest in the package.

The approach of DIME is somewhat less fancy than Turing.jl. Turing does a lot of things for you under the hood. This is nice if things work out well.

DIME is far more conservative. LogProb is assumed to simply take the state space of the posterior distribution as input and to return the posterior density. IMHO, expressing the large-scale time series models used in economics in the "Turing" way would be quite challenging.

Cheers,
Gregor