cv_trac: Perform cross validation for tuning parameter selection
In jacobbien/trac: Tree-based Aggregation of Compositional Data

This function is to be called after calling trac. It performs nfold-fold cross validation.

cv_trac(
  fit,
  Z,
  y,
  A,
  folds = NULL,
  nfolds = 5,
  summary_function = stats::median
)

`fit`	output of `trac` function.
`Z, y, A`	same arguments as passed to `trac`
`folds`	a partition of `1:nrow(Z)`.
`nfolds`	number of folds for cross-validation
`summary_function`	how to combine the errors calculated on each observation within a fold (e.g. mean or median)