assess: assess a neural estimator

In NeuralEstimators: Likelihood-Free Parameter Estimation using Neural Networks

assess a neural estimator

Description

assess a neural estimator

Usage

assess(estimator, parameters, Z, ...)

Arguments

estimator	a neural estimator (or a list of neural estimators)
parameters	true parameters, stored as a d\times K matrix, where d is the dimension of the parameter vector and K is the number of sampled parameter vectors
Z	data simulated conditionally on the parameters. If length(Z) > K, the parameter matrix will be recycled by horizontal concatenation as parameters = parameters[, rep(1:K, J)], where J = length(Z) / K
...	additional keyword arguments passed to the Julia version of assess()

Value

a list with the following elements:

• estimates: a long-form data frame with columns

  • "parameter"; the name of the parameter

  • "truth"; the true value of the parameter

  • "estimate"; the estimated value of the parameter

  • "k"; the index of the parameter vector in the test set

  • "j"; the index of the data set

  For an IntervalEstimator, estimate is replaced by columns lower and upper. For a QuantileEstimator, there is an additional column prob giving the probability level of each quantile estimate.

• runtimes: a data frame giving the total time taken for each estimator

• samples (posterior and ratio estimators only): posterior samples stored as a long-form Julia DataFrame with columns parameter, truth, k, and j as above, plus draw (the index of the draw) and value (the sampled value). This object is not converted to an R data frame, which is slow for large sample collections; extract or summarise it in Julia if needed.

See Also

risk(), rmse(), bias(), plotestimates(), and plotdistribution() for computing various empirical diagnostics and visualisations from an object returned by assess()

NeuralEstimators documentation built on Sept. 10, 2026, 1:12 a.m.