PCR_ensemble_selection {ldsr}

Select the best reconstruction

Description

Select the best reconstruction from an ensemble. Experimental, API may change.

Usage

PCR_ensemble_selection(
  Qa,
  pc,
  start.year,
  transform = "log",
  Z = NULL,
  agg.type = c("best member", "best overall"),
  criterion = c("RE", "CE", "nRMSE", "KGE"),
  return.all.metrics = FALSE
)

Arguments

Qa

Observations: a data.frame of annual streamflow with at least two columns: year and Qa.

pc

For a single model: a data.frame, one column for each principal component. For an ensemble reconstruction: a list, each element is a data.frame of principal components.

start.year

Starting year of the climate proxies, i.e, the first year of the paleo period. start.year + nrow(pc) - 1 will determine the last year of the study horizon, which must be greater than or equal to the last year in Qa.

transform

Flow transformation, either "log", "boxcox" or "none". Note that if the Box-Cox transform is used, the confidence interval after back-transformation is simply the back-transform of the trained onfidence interval; this is hackish and not entirely accurate.

Z

A list of cross-validation folds. If NULL, will be created with make_Z() with default settings. Users are advised to use make_Z() to create the cross-validation folds beforehand. See make_Z for details.

agg.type

Type of ensemble aggregate. There are 2 options: 'best member': the member with the best performance score is used; 'best overall': if the ensemble average is better than the best member, it will be used, otherwise the best member will be used.

criterion

The performance criterion to be used.

return.all.metrics

Logical, if TRUE, all members' performance scores (and the ensemble average's score, if agg.type == 'best overall') are returned.

Value

A list of two elements:

• choice: The index of the selection. If the ensemble is selected, returns 0.

• cv: the cross-validation results of the choice, see cvPCR for details.

• all.metrics: all members' scores, and if agg.type == 'best overall', the ensemble average's scores as well, in the last column.

Examples

PCR_ensemble_selection(NPannual, list(NPpc, NPpc[, 1:2]), start.year = 1200,
                       agg.type = 'best overall', criterion = 'KGE')
PCR_ensemble_selection(NPannual, list(NPpc, NPpc[, 1:2]), start.year = 1200,
                       agg.type = 'best overall', criterion = 'KGE')