evaluate_CV: Evaluation criteria

In convoSPAT: Convolution-Based Nonstationary Spatial Modeling

Description

Calculate three evaluation criteria – continuous rank probability score (CRPS), prediction mean square deviation ratio (pMSDR), and mean squared prediction error (MSPE) – comparing hold-out data and predictions.

Usage

evaluate_CV(holdout.data, pred.mean, pred.SDs)

Arguments

holdout.data	Observed/true data that has been held out for model comparison.
pred.mean	Predicted mean values corresponding to the hold-out locations.
pred.SDs	Predicted standard errors corresponding to the hold-out locations.

Value

A list with the following components:

CRPS	The CRPS averaged over all hold-out locations.
MSPE	The mean squared prediction error.
pMSDR	The prediction mean square deviation ratio.

Examples

## Not run: 
evaluate_CV( holdout.data = simdata$sim.data[holdout.index],
pred.mean = pred.NS$pred.means, pred.SDs = pred.NS$pred.SDs )

## End(Not run)

convoSPAT documentation built on Jan. 16, 2021, 5:31 p.m.