You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I just noticed that simplify doesn't actually use cabinetry (as I assumed purely looking at the plots) but reuses some of that code with slight modifications and I was curious whether things were missing from the cabinetry API / which modificatoins were necessary.
If @alexander-held agrees maybe it makes sense to try to upstream some of these modifications and then add cabinetry as a dependency
(I could also see an argument of simplify being a pure likelihood translator but the diagnostics seem nice to have)
wow, only took me 2 months to notice this comment ...
You're exactly right, I took some of the cabinetry plotting code for the diagnostics stuff. I'll check how easy/sensible it would be to upstream this to cabinetry.
The argument of understanding simplify to be a simple translator can of course be made, but I always found some minimal diagnostics quite helpful. I didn't plan to have anything fancier than simple pull plots, correlation matrices and yields tables though (so really just to make sure the full likelihood that goes in behaves the way you'd expect it to).
Activity
eschanet commentedon May 26, 2021
Hi @lukasheinrich,
wow, only took me 2 months to notice this comment ...
You're exactly right, I took some of the
cabinetryplotting code for the diagnostics stuff. I'll check how easy/sensible it would be to upstream this tocabinetry.The argument of understanding
simplifyto be a simple translator can of course be made, but I always found some minimal diagnostics quite helpful. I didn't plan to have anything fancier than simple pull plots, correlation matrices and yields tables though (so really just to make sure the full likelihood that goes in behaves the way you'd expect it to).