model is any function model(X_tr, y_tr, X_te) returning predictions, so
this is independent of any particular estimator; the default is OLS with an
intercept. The reported se is across folds, which is what the
one-standard-error rule uses – the folds share training data and so are
positively correlated, biasing it downward.
Usage
morie_esl_cv_score(
X,
y,
model = NULL,
k = 5L,
loss = "mse",
stratify = FALSE,
seed = 0L
)
Arguments
- X
Predictors, n by p.
- y
Response.
- model
Function (X_tr, y_tr, X_te) -> predictions; defaults to OLS.
- k
Number of folds, 2 to n.
- loss
"mse", "mae" or "01".
- stratify
Preserve class proportions per fold.
- seed
Seed for the fold shuffle.
Value
List with cv, se, fold_scores, predictions, fold_id.
References
Hastie, T., et al. (2009). ESL (2nd ed.), Sec 7.10. Springer.
Examples
set.seed(1)
X <- matrix(rnorm(300), ncol = 3)
y <- X %*% c(1, -2, 0.5) + rnorm(100, sd = 0.3)
morie_esl_cv_score(X, y, k = 5)$cv < stats::var(y)
#> [,1]
#> [1,] TRUE
