Energy Bayesian fraction of missing information

Formula: EBFMI = sum_{n=1}^{N} (E_n - E_{n-1})^2 / sum_{n=0}^{N} (E_n - Ebar)^2

Usage

Ebfmi(energy)

Arguments

energy

See Usage.

Value

List with ebfmi (per chain), min_ebfmi, n_chains, n.

References

Betancourt (2016), Diagnosing Suboptimal Cotangent Disintegrations in Hamiltonian Monte Carlo, arXiv:1604.00695. Verified against the paper: the estimator is the displayed equation for BFMI-hat.

Examples

V <- c(1, 2, 3, 4, 5, 6, 7, 8)
Ebfmi(V)
#> $ebfmi
#> [1] 0.1666667
#> 
#> $min_ebfmi
#> [1] 0.1666667
#> 
#> $n_chains
#> [1] 1
#> 
#> $n
#> [1] 8
#> 
#> $method
#> [1] "Energy Bayesian fraction of missing information"
#> 
#> attr(,"class")
#> [1] "morie_rich_result" "list"