Fit

Created on Fri May 24 09:38 2024

@author: MCR

Functions for fitting light curve models to data.

class exouprf.fit.Dataset(input_parameters, t, lc_model_type, linear_regressors=None, observations=None, gp_regressors=None, ld_model='quadratic', silent=False, custom_lc_functions=None)

Primary exoUPRF class. Stores a set of light curve observations and performs light curve fits.

fit(output_file, sampler='MCMC', mcmc_start=None, mcmc_ncores=1, mcmc_steps=10000, continue_mcmc=False, dynesty_args=None, force_redo=False, resume_dynesty=False, dynesty_resume_file=None)

Run a light curve fit.

Parameters:

• output_file (str) – Path to file to which to save outputs.

• sampler (str) – Sampling to use, either ‘MCMC’ or ‘Nested Sampling’.

• mcmc_start (ndarray(float)) – Starting positions for MCMC sampling. MCMC only.

• mcmc_steps (int) – Number of steps to take for MCMC sampling. MCMC only.

• mcmc_ncores (int) – Number of cores for multiprocessing. MCMC only.

• continue_mcmc (bool) – If True, continue from a previous MCMC run saved in output_file. MCMC only.

• dynesty_args (dict) – Keyword arguments to pass to the dynesty NestedSampler instance. Nested Sampling only.

• force_redo (bool) – If True, will overwrite previous output files.

• resume_dynesty (bool) – If True, restart sampling from a previous fit.

• dynesty_resume_file (str, None) – Previous dynesty sampling save file from which to resume the fit.

get_param_dict_from_fit(method='median', mcmc_burnin=None, mcmc_thin=15, drop_chains=None)

Reformat MCMC fit outputs into the parameter dictionary format expected by Model.

Parameters:

• method (str) – Method via which to get best fitting parameters from MCMC chains. Either “median” or “maxlike”.

• mcmc_burnin (int) – Number of steps to discard as burn in. Defaults to 75% of chain length. Only for MCMC.

• mcmc_thin (int) – Increment by which to thin chains. Only for MCMC.

• drop_chains (list(int), None) – Indices of chains to drop.

Returns:

param_dict – Dictionary of light curve model parameters.

Return type:

dict

get_results_from_fit(mcmc_burnin=None, mcmc_thin=15, drop_chains=None)

Extract MCMC posterior sample statistics (median and 1 sigma bounds) for each fitted parameter.

Parameters:

• mcmc_burnin (int) – Number of steps to discard as burn in. Defaults to 75% of chain length.

• mcmc_thin (int) – Increment by which to thin chains.

• drop_chains (list(int), None) – Indices of chains to drop.

Returns:

results_dict – Dictionary of posterior medians and 1 sigma bounds for each fitted parameter.

Return type:

dict

make_corner_plot(mcmc_burnin=None, mcmc_thin=15, labels=None, outpdf=None, log_params=None, drop_chains=None)

Make a corner plot of fitted posterior distributions.

Parameters:

• mcmc_burnin (int) – Number of steps to discard as burn in. Defaults to 75% of chain length.

• mcmc_thin (int) – Increment by which to thin chains.

• labels (list(str)) – Fitted parameter names.

• outpdf (PdfPages) – File to save plot.

• log_params (list(int), None) – Indices of parameters to plot in log-space.

• drop_chains (list(int), None) – Indices of chains to drop.

plot_mcmc_chains(labels=None, log_params=None, highlight_chains=None, drop_chains=None)

Plot MCMC chains.

Parameters:

• labels (list(str)) – Fitted parameter names.

• log_params (list(int), None) – Indices of parameters to plot in log-space.

• highlight_chains (list(int), None) – Indices of chains to highlight.

• drop_chains (list(int), None) – Indices of chains to drop.

exouprf.fit.fit_dynesty(prior_transform, log_like, ndim, output_file, log_like_args, ptform_kwargs, dynesty_args=None, silent=False, resume=False, resume_file=None)

Run a light curve fit via nested sampling using the dynesty.

Parameters:

• prior_transform (function) – Callable function to evaluate the prior transform.

• log_like (function) – Callable function to evaluate the log likelihood.

• ndim (int) – Number of sampling dimensions.

• output_file (str) – File to which to save outputs.

• log_like_args (dict) – Arguments for the log likelihood function.

• ptform_kwargs (dict) – Arguments for the prior transform function.

• dynesty_args (dict) – Arguments for dynesty NestedSampler instance.

• silent (bool) – If True, do not show progress updates.

• resume (bool) – If True, restart sampling from a previous fit.

• resume_file (str, None) – Previous dynesty save file to restart from.

Returns:

sampler – dynesty sampler.

Return type:

dynesty.nestedsamplers.MultiEllipsoidSampler

exouprf.fit.fit_emcee(log_prob, output_file, initial_pos=None, continue_run=False, silent=False, mcmc_steps=10000, log_probability_args=None, ncores=1)

Run a light curve fit via MCMC using the emcee sampler.

Parameters:

• log_prob (function) – Callable function to evaluate the fit log probability.

• output_file (str) – File to which to save outputs. If continuing a run, this should also be the input file containing the previous MCMC chains.

• initial_pos (ndarray(float), None) – Starting positions for the MCMC sampling.

• continue_run (bool) – If True, continue a run from the state of previous MCMC chains.

• silent (bool) – If True, do not show any progress.

• mcmc_steps (int) – Number of MCMC steps before stopping.

• log_probability_args (tuple) – Arguments for the passed log_prob function.

• ncores (int) – Number of cores to use for multiprocessing.

Returns:

sampler – ecmee sampler.

Return type:

emcee.ensemble.EnsembleSampler

exouprf.fit.log_likelihood(theta, param_dict, time, observations, mod_init)

Evaluate the log likelihood for a dataset and a given set of model parameters.

Parameters:

• theta (list(float)) – List of values for each fitted parameter.

• param_dict (dict) – Dictionary of input parameter values and prior distributions.

• time (dict) – Dictonary of timestamps corresponding to the observations.

• observations (dict) – Dictionary of observations.

• mod_init (LightCurveModel) – Initialized LightCurveModel class.

Returns:

log_like – Result of likelihood evaluation.

Return type:

float

exouprf.fit.log_probability(theta, param_dict, time, observations, mod_init)

Evaluate the log probability for a dataset and a given set of model parameters.

Parameters:

• theta (list(float)) – List of values for each fitted parameter.

• param_dict (dict) – Dictionary of input parameter values and prior distributions.

• time (dict) – Dictonary of timestamps corresponding to the observations.

• observations (dict) – Dictionary of observations.

• mod_init (LightCurveModel) – Initialized LightCurveModel class.

Returns:

log_prob – Result of probability evaluation.

Return type:

float

exouprf.fit.logprior_loguniform(x, hyperparams)

Evaluate log-uniform log prior.

exouprf.fit.logprior_normal(x, hyperparams)

Evaluate normal log prior.

exouprf.fit.logprior_truncatednormal(x, hyperparams)

Evaluate trunctaed normal log prior.

exouprf.fit.logprior_uniform(x, hyperparams)

Evaluate uniform log prior.

exouprf.fit.set_logprior(theta, param_dict)

Calculate the fit prior based on a set of input values and prior functions.

Parameters:

• theta (list(float)) – List of values for each fitted parameter.

• param_dict (dict) – Dictionary of input parameter values and prior distributions.

Returns:

log_prior – Result of prior evaluation.

Return type:

float

exouprf.fit.set_prior_transform(theta, param_dict)

Define the prior transform based on a set of input values and prior functions.

Parameters:

• theta (list(float)) – List of values for each fitted parameter.

• param_dict (dict) – Dictionary of input parameter values and prior distributions.

Returns:

prior_transform – Result of prior evaluation.

Return type:

list(float)

exouprf.fit.transform_loguniform(x, hyperparams)

Evaluate log-uniform prior transform.

exouprf.fit.transform_normal(x, hyperparams)

Evaluate normal prior transform.

exouprf.fit.transform_truncatednormal(x, hyperparams)

Evaluate truncated normal prior transform.

exouprf.fit.transform_uniform(x, hyperparams)

Evaluate uniform prior transform.