pymcmcstat.structures package

pymcmcstat.structures.ParameterSet module

Created on Thu Jan 18 10:15:37 2018

@author: prmiles

class pymcmcstat.structures.ParameterSet.ParameterSet(theta=None, ss=None, prior=None, sigma2=None, alpha=None)

Bases: object

Basic MCMC parameter set.

Description: Storage device for passing parameter sets back and forth between sampling methods.

Args:

• theta (ndarray): Sampled values.

• ss (ndarray): Sum-of-squares error(s).

• prior (ndarray): Result from prior function.

• sigma2 (ndarray): Observation errors.

• alpha (float): Result from evaluating likelihood function.

pymcmcstat.structures.ResultsStructure module

Created on Wed Jan 17 09:18:19 2018

@author: prmiles

class pymcmcstat.structures.ResultsStructure.ResultsStructure

Bases: object

Results from MCMC simulation.

Description: Class used to organize results of MCMC simulation.

Attributes:

• export_simulation_results_to_json_file()

• determine_filename()

• save_json_object()

• load_json_object()

• add_basic()

• add_updatesigma()

• add_dram()

• add_prior()

• add_options()

• add_model()

• add_chain()

• add_s2chain()

• add_sschain()

• add_time_stats()

• add_random_number_sequence()

add_basic(nsimu, covariance, parameters, rejected, simutime, theta)

Add basic results from MCMC simulation to structure.

Args:

• nsimu (int): Number of MCMC simulations.

• model (ModelSettings): MCMC model settings.

• covariance (CovarianceProcedures): Covariance variables.

• parameters (ModelParameters): Model parameters.

• rejected (dict): Dictionary of rejection stats.

• simutime (float): Simulation run time in seconds.

• theta (ndarray): Last sampled values.

add_chain(chain=None)

Add chain to results structure.

Args:

• chain (ndarray): Model parameter sampling chain.

add_dram(drscale, RDR, total_rejected, drsettings)

Add results specific to performing DR algorithm.

Args:

• drscale (ndarray): Reduced scale for sampling in DR algorithm. Default is [5,4,3].

• RDR (ndarray): Cholesky decomposition of covariance matrix based on DR.

• total_rejected (int): Number of rejected samples.

• drsettings (DelayedRejection): Need access to counters within DR class.

add_model(model=None)

Saves subset of features of the model settings in a nested dictionary.

Args:

• model (ModelSettings): MCMC model settings.

add_options(options=None)

Saves subset of features of the simulation options in a nested dictionary.

Args:

• options (SimulationOptions): MCMC simulation options.

add_prior(mu, sigma, priortype)

Add results specific to prior function.

Args:

• mu (ndarray): Prior mean.

• sigma (ndarray): Prior standard deviation.

• priortype (int): Flag identifying type of prior.

Note

This feature is not currently implemented.

add_random_number_sequence(rndseq)

Add random number sequence to results structure.

Args:

• rndseq (ndarray): Sequence of sampled random numbers.

Note

This feature is not currently implemented.

add_s2chain(s2chain=None)

Add observiation error chain to results structure.

Args:

• s2chain (ndarray): Sampling chain of observation errors.

add_sschain(sschain=None)

Add sum-of-squares chain to results structure.

Args:

• sschain (ndarray): Calculated sum-of-squares error for each parameter chains set.

add_time_stats(mtime, drtime, adtime)

Add time spend using each sampling algorithm.

Args:

• mtime (float): Time spent performing standard Metropolis.

• drtime (float): Time spent performing Delayed Rejection.

• adtime (float): Time spent performing Adaptation.

Note

This feature is not currently implemented.

add_updatesigma(updatesigma, sigma2, S20, N0)

Add information to results structure related to observation error.

Args:

• updatesigma (bool): Flag to update error variance(s).

• sigma2 (ndarray): Latest estimate of error variance(s).

• S20 (ndarray): Scaling parameter(s).

• N0 (ndarray): Shape parameter(s).

If updatesigma is True, then

results['sigma2'] = np.nan
results['S20'] = S20
results['N0'] = N0

Otherwise

results['sigma2'] = sigma2
results['S20'] = np.nan
results['N0'] = np.nan

classmethod determine_filename(options)

Determine results filename.

If not specified by results_filename in the simulation options, then a default naming format is generated using the date string associated with the initialization of the simulation.

Args:

• options (SimulationOptions): MCMC simulation options.

Returns:

• filename (str): Filename string.

export_lightly(results)

Export minimal simulation results to a json file.

This will save the key terms in the results dict, excluding arrays. Ideally, this is used in conjunction with one of the chain saving methods. The goal is to provide a results dict to simplify post- processing and reduces storage overhead.

Args:

• results (ResultsStructure): Dictionary of MCMC simulation results/settings.

export_simulation_results_to_json_file(results)

Export simulation results to a json file.

Args:

• results (ResultsStructure): Dictionary of MCMC simulation results/settings.

classmethod load_json_object(filename)

Load object stored in json file.

Note

Filename should include extension.

Args:

• filename (str): Load object from file with this name.

Returns:

• results (dict): Object loaded from file.

classmethod save_json_object(results, filename)

Save object to json file.

Note

Filename should include extension.

Args:

• results (dict): Object to save.

• filename (str): Write object into file with this name.

pymcmcstat.structures.ResultsStructure.lighten_results(results)

Saves subset of features of the simulation options in a nested dictionary.

Args:

• options (SimulationOptions): MCMC simulation options.