Incomplete signal patch treatment

[jackaraz]

This PR improves patch set handling when patches do not include any information about certain channels.

Thanks to @Rav2 for pointing this out.

[Rav2]

TEST 1

1. Load data from ATLAS 1911.12606 analysis for EWKinos
2. Create wrapper with the background and the first patch
3. Get model and data for the 'apriori' ExpectationType
4. The output:

Spey - WARNING: Some of the channels are not defined in the patch set, these channels will be kept in the statistical model. 
Spey - WARNING: If these channels are meant to be removed, please indicate them in the patch set.
Spey - WARNING: Please check the following channel(s): SR_eMLLa_Onelep1track_cuts, SR_eMLLb_Onelep1track_cuts, SR_eMLLc_Onelep1track_cuts, SR_eMLLd_Onelep1track_cuts, SR_eMLLe_Onelep1track_cuts, SR_eMLLf_Onelep1track_cuts, SRee_eMLLf_lowmet_deltaM_low_cuts, SRmm_eMLLh_lowmet_deltaM_high_cuts

TEST2

1. Use the model from TEST1
2. Calculate nLLs for ExpectationTypes: ['apriori', 'aposteriori', 'observed'] for mu=0 and mu=1
3. Output:

spey (apriori): nLL_exp_mu0=258.82 nLL_exp_mu1=265.83
spey (aposteriori): nLL_exp_mu0=294.44 nLL_exp_mu1=296.38
spey: nLL_obs_mu0=294.44 nLL_obs_mu1=296.38

TEST3

1. Repeat TEST1 for all files and patches from 1911.12606 analysis
2. Observation: sometimes many SRs are missing in the patches
3. Result: PASSED

COMMENTS AND SUGGESTIONS

1. Overall, it works good.
2. The warning messages are good, but wouldn't it be better to merge them into one, or two (first merged with second) rather than print 3?
3. When user loads multiple patches, like I do in TEST3, or when we want to calculate limit, it would be helpful to print the name of the patch in the warning (a check would be needed if there is a name in the metadata), because different patches (mass points) might be missing information for various channels.
4. An optional flag would be nice to silence these warnings if we know what we are doing.

[jackaraz]

TEST 1

1. Load data from ATLAS 1911.12606 analysis for EWKinos
2. Create wrapper with the background and the first patch
3. Get model and data for the 'apriori' ExpectationType
4. The output:

Spey - WARNING: Some of the channels are not defined in the patch set, these channels will be kept in the statistical model. 
Spey - WARNING: If these channels are meant to be removed, please indicate them in the patch set.
Spey - WARNING: Please check the following channel(s): SR_eMLLa_Onelep1track_cuts, SR_eMLLb_Onelep1track_cuts, SR_eMLLc_Onelep1track_cuts, SR_eMLLd_Onelep1track_cuts, SR_eMLLe_Onelep1track_cuts, SR_eMLLf_Onelep1track_cuts, SRee_eMLLf_lowmet_deltaM_low_cuts, SRmm_eMLLh_lowmet_deltaM_high_cuts

TEST2

1. Use the model from TEST1
2. Calculate nLLs for ExpectationTypes: ['apriori', 'aposteriori', 'observed'] for mu=0 and mu=1
3. Output:

spey (apriori): nLL_exp_mu0=258.82 nLL_exp_mu1=265.83
spey (aposteriori): nLL_exp_mu0=294.44 nLL_exp_mu1=296.38
spey: nLL_obs_mu0=294.44 nLL_obs_mu1=296.38

TEST3

1. Repeat TEST1 for all files and patches from 1911.12606 analysis
2. Observation: sometimes many SRs are missing in the patches
3. Result: PASSED

COMMENTS AND SUGGESTIONS

The warning messages are good, but wouldn't it be better to merge them into one or two (first merged with second) rather than print 3?

That's intentional, actually; otherwise, everything gets into each other and makes things avoidable. I want it to scream to the user.

An optional flag would be nice to silence these warnings if we know what we are doing.

That is already there see the documentation.

When user loads multiple patches, like I do in TEST3, or when we want to calculate limit, it would be helpful to print the name of the patch in the warning (a check would be needed if there is a name in the metadata), because different patches (mass points) might be missing information for various channels.

Patch sets does not necessarily have a unique name, it will work even if you give the same name to all so that wont change anything.

[Rav2]

Seal of approval.