Parameter bounds affect fits with fixed parameters

Description

See this example:

import pyhf

spec = {
    "channels": [
        {
            "name": "SR",
            "samples": [
                {
                    "data": [10.0],
                    "modifiers": [
                        {"data": None, "name": "Signal_norm", "type": "normfactor"}
                    ],
                    "name": "Signal",
                },
                {"data": [50.0], "modifiers": [], "name": "Background"},
            ],
        }
    ],
    "measurements": [
        {
            "config": {
                "parameters": [
                    {"bounds": [[0, 5]], "inits": [2.0], "name": "Signal_norm"}
                ],
                "poi": "Signal_norm",
            },
            "name": "minimal_example",
        }
    ],
    "observations": [{"data": [65.0], "name": "SR"}],
    "version": "1.0.0",
}

ws = pyhf.Workspace(spec)
model = ws.model()
data = ws.data(model)

# pyhf.infer.hypotest(6.0, data, model, qtilde=True)  # this fails

pyhf.set_backend("numpy", pyhf.optimize.minuit_optimizer(verbose=False))
# the next two work, but result in identical values
print(pyhf.infer.hypotest(6.0, data, model, qtilde=True))
print(pyhf.infer.hypotest(7.0, data, model, qtilde=True))

The CLs value reported in the last two lines is identical. This seems to be related to the parameter bounds [0,5]. When increasing the range, the CLs values are different (as expected).

When uncommenting the first hypotest (using scipy), the minimization fails:

pyhf.exceptions.FailedMinimization: Positive directional derivative for linesearch

Expected Behavior

All fits succeed, and different CLs for any POI value. If the bounds are somehow still respected, a warning about parameters being outside the bounds (and scaled back to the bound) would be helpful.

Actual Behavior

Same CLs for POI values above bound, and scipy backend crashes.

Steps to Reproduce

See above, same behavior with 0.5.2 and master.

Checklist

• Run git fetch to get the most up to date version of master
• Searched through existing Issues to confirm this is not a duplicate issue
• Filled out the Description, Expected Behavior, Actual Behavior, and Steps to Reproduce sections above or have edited/removed them in a way that fully describes the issue

[lukasheinrich]

I’ve verified that the example from @alexander-held fails with a more interpretable error msg for all three cases in #1126 and once you adjust the bounds the CLs values behave as expected

[matthewfeickert]

@lukasheinrich this is great. I think we can say that PR #1126 fully addresses this Issue then.