EstimateResult

Note

A call of an IBS object with additional_output="full" returns an EstimateResult: the estimate, its variance and standard deviation, the exit flag and its message, the cost of the call, and the estimate of each trial, with its variance.

It is a dictionary whose keys can also be read as attributes: result["neg_logl"] and result.neg_logl are the same value.

class pyibs.EstimateResult

The result of an IBS call with additional_output="full".

A dictionary whose keys can also be read as attributes.

Attributes:

neg_loglfloat

The negative log-likelihood estimate, or the log-likelihood estimate with return_positive=True.

neg_logl_varfloat

Estimated variance of neg_logl.

neg_logl_stdfloat

Estimated standard deviation of neg_logl: the square root of neg_logl_var.

exit_flagint

0 if all requested repeats completed without reaching the likelihood threshold; 1 if the threshold ended at least one repeat; 2 if max_time stopped the sampling. Flag 2 takes precedence when both stopping rules apply. These flags describe how the call ended; see the stopping settings of IBS for their statistical consequences.

messagestr

The exit flag’s meaning.

elapsed_timefloat

Seconds spent in the call.

num_samples_per_trialfloat

Total simulated responses divided by the number of trials. Includes samples drawn after a trial’s last required match.

fun_countint

Number of simulator calls used to produce this estimate.

neg_logl_trialsndarray of shape (N,)

Unweighted negative log-likelihood estimate for each trial, averaged over its completed repeats. All entries are NaN if the likelihood threshold ended any repeat. Otherwise, their weighted sum equals neg_logl up to rounding, with the opposite sign when return_positive=True.

neg_logl_var_trialsndarray of shape (N,)

Estimated variance for each entry of neg_logl_trials. All entries are NaN if the threshold ended any repeat. Otherwise, summing them with squared trial weights gives neg_logl_var.