Model-Evaluation Metrics

Metrics used during model evaluation and model comparison

Module Documentation

fitr.criticism.model_evaluation.AIC(nparams, loglik)

Calculates Aikake information criterion

Parameters:

nparams : int

Number of parameters in the model

loglik : float or ndarray(dtype=float)

Log-likelihood

Returns:

float or ndarray(dtype=float)

fitr.criticism.model_evaluation.BIC(loglik, nparams, nsteps)

Calculates Bayesian information criterion

Parameters:

loglik : float or ndarray(dtype=float)

Log-likelihood

nparams : int

Number of parameters in the model

nsteps : int

Number of time steps in the task

Returns:

float or ndarray(dtype=float)

fitr.criticism.model_evaluation.LME(logpost, nparams, hessian)

Laplace approximated log-model-evidence (LME)

Parameters:

logpost : float or ndarray(dtype=float)

Log-posterior probability

nparams : int

Number of parameters in the model

hessian : ndarray(size=(nparams, nparams))

Hessian computed from parameter optimization

Returns:

float or ndarray(dtype=float)