Fit#

class gammapy.modeling.Fit(backend='minuit', optimize_opts=None, covariance_opts=None, confidence_opts=None, store_trace=False)[source]#

Bases: object

Fit class.

The fit class provides a uniform interface to multiple fitting backends. Currently available: “minuit”, “sherpa” and “scipy”.

Parameters

backend{“minuit”, “scipy” “sherpa”}

Global backend used for fitting. Default is “minuit”.

optimize_optsdict

Keyword arguments passed to the optimizer. For the "minuit" backend see https://iminuit.readthedocs.io/en/stable/reference.html#iminuit.Minuit for a detailed description of the available options. If there is an entry ‘migrad_opts’, those options will be passed to iminuit.Minuit.migrad().

For the "sherpa" backend you can from the options:

"simplex"

"levmar"

"moncar"

"gridsearch"

Those methods are described and compared in detail on http://cxc.cfa.harvard.edu/sherpa/methods/index.html. The available options of the optimization methods are described on the following pages in detail:

http://cxc.cfa.harvard.edu/sherpa/ahelp/neldermead.html

http://cxc.cfa.harvard.edu/sherpa/ahelp/montecarlo.html

http://cxc.cfa.harvard.edu/sherpa/ahelp/gridsearch.html

http://cxc.cfa.harvard.edu/sherpa/ahelp/levmar.html

For the "scipy" backend the available options are described in detail here: https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html

covariance_optsdict

Covariance options passed to the given backend.

confidence_optsdict

Extra arguments passed to the backend. E.g. iminuit.Minuit.minos supports a maxcall option. For the scipy backend confidence_opts are forwarded to brentq. If the confidence estimation fails, the bracketing interval can be adapted by modifying the upper bound of the interval (b) value.

store_tracebool

Whether to store the trace of the fit.

Methods Summary

`confidence`(datasets, parameter[, sigma, ...])	Estimate confidence interval.
`covariance`(datasets[, optimize_result])	Estimate the covariance matrix.
`optimize`(datasets)	Run the optimization.
`run`(datasets)	Run all fitting steps.
`stat_contour`(datasets, x, y[, numpoints, sigma])	Compute stat contour.
`stat_profile`(datasets, parameter[, reoptimize])	Compute fit statistic profile.
`stat_surface`(datasets, x, y[, reoptimize])	Compute fit statistic surface.

Methods Documentation

confidence(datasets, parameter, sigma=1, reoptimize=True)[source]#

Estimate confidence interval.

Extra kwargs are passed to the backend. E.g. iminuit.Minuit.minos supports a maxcall option.

For the scipy backend kwargs are forwarded to brentq. If the confidence estimation fails, the bracketing interval can be adapted by modifying the the upper bound of the interval (b) value.

Parameters

datasetsDatasets or list of Dataset: Datasets to optimize.
parameterParameter: Parameter of interest.
sigmafloat, optional: Number of standard deviations for the confidence level. Default is 1.
reoptimizebool, optional: Re-optimize other parameters, when computing the confidence region. Default is True.

Returns

resultdict: Dictionary with keys “errp”, ‘errn”, “success” and “nfev”.

covariance(datasets, optimize_result=None)[source]#

Estimate the covariance matrix.

Assumes that the model parameters are already optimised.

Parameters

datasetsDatasets or list of Dataset: Datasets to optimize.
optimize_resultOptimizeResult, optional: Optimization result. Can be optionally used to pass the state of the IMinuit object to the covariance estimation. This might save computation time in certain cases. Default is None.

Returns

resultCovarianceResult: Results.

optimize(datasets)[source]#

Run the optimization.

Parameters

datasetsDatasets or list of Dataset: Datasets to optimize.

Returns

optimize_resultOptimizeResult: Optimization result.

run(datasets)[source]#

Run all fitting steps.

Parameters

datasetsDatasets or list of Dataset: Datasets to optimize.

Returns

fit_resultFitResult: Fit result.

stat_contour(datasets, x, y, numpoints=10, sigma=1)[source]#

Compute stat contour.

Calls iminuit.Minuit.mncontour.

This is a contouring algorithm for a 2D function which is not simply the fit statistic function. That 2D function is given at each point (par_1, par_2) by re-optimising all other free parameters, and taking the fit statistic at that point.

Very compute-intensive and slow.

Parameters

datasetsDatasets or list of Dataset: Datasets to optimize.
x, yParameter: Parameters of interest.
numpointsint, optional: Number of contour points. Default is 10.
sigmafloat, optional: Number of standard deviations for the confidence level. Default is 1.

Returns

resultdict: Dictionary containing the parameter values defining the contour, with the boolean flag “success” and the information objects from mncontour.

stat_profile(datasets, parameter, reoptimize=False)[source]#

Compute fit statistic profile.

The method used is to vary one parameter, keeping all others fixed. So this is taking a “slice” or “scan” of the fit statistic.

Parameters

datasetsDatasets or list of Dataset: Datasets to optimize.
parameterParameter: Parameter of interest. The specification for the scan, such as bounds and number of values is taken from the parameter object.
reoptimizebool, optional: Re-optimize other parameters, when computing the confidence region. Default is False.

Returns

resultsdict: Dictionary with keys “parameter_name_scan”, “stat_scan” and “fit_results”. The latter contains an empty list, if reoptimize is set to False.

Notes

The progress bar can be displayed for this function.

stat_surface(datasets, x, y, reoptimize=False)[source]#

Compute fit statistic surface.

The method used is to vary two parameters, keeping all others fixed. So this is taking a “slice” or “scan” of the fit statistic.

Caveat: This method can be very computationally intensive and slow