< Mujpy.InstallWin10 | Index | Mujpy.GlobalFits >
New login suggested by PB, in order to allow command line driving, musrfit-style. See Mujpy.GlobalFits for the A1, ... , C2 types of fit.
The mudash class is a separate second layer (relying on ipywidget, jupyterlab, to be done). Several use modes can be envisaged, starting from a python script. Tests of mujpy are provided as jupyter notebook, .ipynb files in example. The core mujpy is defined from the following classes:
- the musuite class, contains everything about data. Use:
- Define an instance. the instance methods that
- defines a console, according to mujpy mode (jupyter, python, gui)
- contains a list of lists of runs, if the innermost contains more items they identify runs to be added. The list defines the fit order for sequentila fits
- reads data for each run from a file, the reference run, only the first full path is given, the others are retrieved from the list
- loads a json
groups.calib file, containing one or more groups, each defined by a dict of two groupings in shorthand notation and an α parameter = {'forward':forw,'backward':backw,'alpha':alpha}, as determined by the special instance. This file can be easily typed in, as in the test notebooks.
- automatic
t0 and background determination, prompt and preprompt, respectively, for musr2py PSI data, and just edge for musr2isis homebrew Nexus ISIS data
- a switchyard asymmetry (distinct methods for Mujpy.A1, Mujpy.A2, Mujpy.B1, Mujpy.B2?, Mujpy.C1, Mujpy.C2).
- the mufit class, contains everything about fitting. It relies on a json structure that reproduces the old mugui dashboard. Use a json aware editor.
- It reads:
- a suite instance (data)
- a json structure defining the model (the name is the game), e.g. 'mgml'. The letters are combinations of 2-letter standard component names, from class
mumodel in library mucomponents. The json file name also refer to the reference run, and the group shorthand
- the combination of the suite and the model determine which type of fit, single or multi run, single or multi group, sequential or global.
- models starting with component "al", for α, produce a special fit, where asymmetry is recalculated at each iteration with the α Minuit parameter; this is the preliminary fit to determine α for the
groups.calib file of the suite.
- the standard model is defined by a structure initialised as the
"model_guess" key of a dict, corresponding to a list of components, that reproduces the old mugui dashboard; here a simple example
[
{"name": "mg",
"label": "_fast"
"pardicts": [
{"name":"A","value":0.2,"error":0.001,"flag":"~","function":"","limits":[null,null]},
{"name":"B","value":10.0,"error":0.1,"flag":"~","function":"","limits":[null,null]},
{"name":'φ',"value":0.0,"error":0.1,"flag":"~","function":"","limits":[null,null]},
{"name":'σ1',"value":3.0,"error":0.1,"flag":"~","function":"","limits":[null,null]}
]
},
{"name": "ml",
"label": "_slow",
"pardicts": [
{"name":"A","value":0.05,"error":0.001,"flag":"~","function":"","limits":[null,null]},
{'name':'B1',"value":0.0,"error":0.001,"flag":"=","function":"p[1]","limits":[null,null]},
{'name':'φ',"value":0.0,"error":0.001,"flag":"=","function":"p[2]","limits":[null,null]},
{'name':'σ1',"value":0.1,"errore":0.001,"flag":"~","function":"","limits":[null,null]},
}
]
- multi group fits may use the additional pardict key "function_multi" and "error_propagation_multi" to define a list of parameter functions and error propagation rules
- global fits have an additional key "userpardicts_guess", assigned to a list of pardicts to define global Minuit parameters, common to suites of runs or to different groupings or both, e.g.
[
{'name':'Atot21','value':0.25,'flag':'~','limits':(None,None)},
{'name':'Atot34','value':0.26,'flag':'~','limits':(None,None)}
]
- mufit methods automatically perform
- Minuit (iminuit2) best fit
- save one or more fit files by adding a
"model_result" copy of the "model_guess" key, with best fit "values" and "std"; global fits save also a "userpardicts_result" key.
- write [in] a csv
- write results on the suite console
- the mufitplot class, produces plots of best/guess fit or fft, or both
- single run, single group fits produce static plots
- all the others produce animations that cycle with 1 s delay all the data sets of the suite. Mouse click can toggle the animation on/off, i.e. stop on any run/group for static observation
- the best fit(s) can be plotted either on a single range on twin ranges with distinct packing: early and late times; chi2 statistics is also shown
- a skeleton rotating frame is selectable, by the rotating frame frequency {$\nu_{\mathrm R}$}, as a presentation option; it works when the fit components are {$\nu-\nu_{\mathrm R}>0$} and {$\nu-\nu_{\mathrm R}\ll\nu_{\mathrm R}$}
- fft is performed on (partial) residues
- the fit component may be subtracted by setting a
"fft":True key in its "mu model_guess"
- a multiplot mode is foreseen, with a vertical offset for clarity.
- a detector inspection mode is foreseen, with animation for scanning detectors
[1] could be a list of lists of lists with a list of successive calibrations (if, e.g., the calibration dependes on T ranges)
< Mujpy.InstallWin10 | Index | Mujpy.GlobalFits >
