EmceeFitter

Function optimisation interface via MCMC sampling and Chi-squared minimisation

Simple module containing a Python 3 interface to perform function optimisation by minimising the Chi-squared, given some observational data.

The Chi-squared function is sampled using the Affine Invariant MCMC sampler contained in the emcee package.
Subsequent minimisation of the Chi-squared over all possible samples returns the least squares combination of parameters.

Characteristics

Fitter base class

Meant for

• initialising default parameters for the emcee sampler
• defining the (log)prior density distribution function (default='uniform')

Input arguments

• ranges : dictionary consisting of parameter names as keys, and tuple of numerical ranges allowed as items
• priors : categorical name of prior density distribution function (default=uniform, available log)
• walkers : number of walkers in the Affine Invariant sampler (default=100)
• threads : number of threads to open to parallelise the computations (default=None)
• burnin : number of burn-in steps to perform (default=None)
• steps : number of steps to perform in the Affine Invariant sampling

EmceeChi2Fitter main class

Meant to perform the function optimisation
The instance is callable performing the Affine Invariant sampling of the Chi-squared.
Input arguments

• function : function to fit
• xobs : abscissa of the observational points
• yobs : ordinate of the observational points
• errors : uncertainties in the observational values

Output values

• values : least squares parameters array
• samples : array containing all parameters values probed
• lnprobs : array containing all Chi-squared values corresponding to samples coordinates