Alexander Held
Alexander Held
The `pyhf digest` approach does not include interpcodes, right? The reason for only hashing the model is that in this use case of yield uncertainty calculation the measurement does not...
> I think only lumi is the only special one that is set through a measurement object which can change expected data I gave this a try and did not...
When comparing models and checking for equality via model spec + interpcodes, one thing that will of course be missed is different measurement config related things like `model.config.suggested_init()`. Despite these...
I was wondering whether this is related to that issue. It seems a bit different to me, as I understand why it would make sense to set modifiers for bins...
@kratsg I agree, I have thought about this a bit more and there is nothing else I can think of which this API should do in addition in such cases....
[`PrintXML`](https://root.cern.ch/doc/master/classRooStats_1_1HistFactory_1_1Channel.html#ae83497da2ea795d2ed1063e7d43b1ea2) allows using relative paths too. The downside of using relative paths there is that you (currently) need to call `pyhf xml2json` from the same place where the relative paths...
@kratsg pointed out that `ROOT` 6.22 switches the `HistFactory` default constraint term to Poisson: https://root.cern/doc/v622/release-notes.html#release-6.2202
related: #662
Similarly to stat errors, `ShapeSys` has a `ConstraintType` option in HistFactory to switch to a Gaussian term. Presumably that information also is lost in a roundtrip.
Hi @balunas, I can't comment on the time line, but maybe a few comments about the other points: I think the use of Gaussian as default came from the HistFactory...