Bias-variance decomposition (Geron Ch 1, morie.fn hmbv)

Variance across models is the POPULATION variance (ddof = 0); the identity only closes with the biased estimator.

Usage

morie_geron_bias_variance_tradeoff(preds, y, f_true = NULL)

Arguments

preds

Matrix (n_models, n_points).

y

Targets.

f_true

Optional noise-free values.

Value

List with bias2, variance, noise, mse, mean_pred.

Examples

V <- c(1, 2, 3, 4, 5, 6, 7, 8)
morie_geron_bias_variance_tradeoff(V, V)
#> $bias2
#> [1] 0
#> 
#> $variance
#> [1] 0
#> 
#> $noise
#> [1] 0
#> 
#> $mse
#> [1] 0
#> 
#> $mean_pred
#> [1] 1 2 3 4 5 6 7 8
#> 
#> $var_point
#> [1] 0 0 0 0 0 0 0 0
#> 
#> $n_models
#> [1] 1
#> 
#> $estimate
#> [1] 0
#> 
#> $n
#> [1] 8
#> 
#> $method
#> [1] "Bias-variance decomposition E[err] = bias^2 + variance + noise"
#>