Using cross-validation to assess parameter values

Cross-validation is performed automatically when you use the Geostatistical Wizard to build an interpolation model. The results are presented in the last step of the wizard. Cross-validation can also be performed manually using the Cross Validation geoprocessing tool.

The objective of performing cross-validation is to determine the quality of a model. The goal is to have standardized mean prediction errors near 0, small root-mean-squared prediction errors, average standard error near root-mean-squared prediction errors, and standardized root-mean-squared prediction errors near 1.

The spread of the points should be as close as possible around the dashed gray line. Look for points that deviate greatly from the line.

The quality of fit of the model is shown graphically on the cross-validation page of the Geostatistical Wizard. To view this, click the Predict, Error, Standardized Error, or QQ Plot tab on the Cross Validation dialog box according to the desired method with which you want to view the results.

Cross-validation results can be investigated for particular points—usually ones that have poor results (large errors). To do this, click the row representing the point of interest in the table at the left of the Cross Validation dialog box. The point becomes highlighted in cyan on the chart.

Verwandte Themen


7/10/2012