Currently, we allow lines to have exponential tails with fractional weight of up to 100% on both the low and high-energy sides of the main Gaussian. But when their sum exceeds 100%, this makes no sense.
Proposal: change the meaning of the parameters, and change one name to hint at the new distinction.
Old way
tail_frac was the fraction of all events that end up in the low-energy tail. Allowed values: [0,1]
tail_frac_hi was the fraction of all events that end up in the high-energy tail. Allowed values: [0,1]
New way
tail_frac is the fraction of all events that end up in either tail. Allowed values: [0,1]
tail_share_hi is the fraction of all tail events that end up in the high-energy tail. Allowed values: [0,1]
The use of “share” instead of its approximate synonym “frac” is deliberate: it hints at the fact that they are semantically different things, even if admittedly the 2 words are kind of synonyms.
It’s possible that this is related to the new problem with errors found in scipy/lmfit (issue #251). But even if it’s not, we should make it impossible for the fitter to explore unphysical values of tail-related parameters and make the proposed change.
Original report by Joseph Fowler (Bitbucket: joe_fowler, ).
Currently, we allow lines to have exponential tails with fractional weight of up to 100% on both the low and high-energy sides of the main Gaussian. But when their sum exceeds 100%, this makes no sense.
Proposal: change the meaning of the parameters, and change one name to hint at the new distinction.
Old way
tail_frac
was the fraction of all events that end up in the low-energy tail. Allowed values: [0,1]tail_frac_hi
was the fraction of all events that end up in the high-energy tail. Allowed values: [0,1]New way
tail_frac
is the fraction of all events that end up in either tail. Allowed values: [0,1]tail_share_hi
is the fraction of all tail events that end up in the high-energy tail. Allowed values: [0,1]The use of “share” instead of its approximate synonym “frac” is deliberate: it hints at the fact that they are semantically different things, even if admittedly the 2 words are kind of synonyms.
It’s possible that this is related to the new problem with errors found in scipy/lmfit (issue #251). But even if it’s not, we should make it impossible for the fitter to explore unphysical values of tail-related parameters and make the proposed change.