Skip to main content

Cross Validation (Geostatistical Analyst Tools)

Summary

Removes one data location and predicts the associated data using the data at the rest of the locations. The primary use for this tool is to compare the predicted value to the observed value in order to obtain useful information about some of your model parameters.

Learn more about performing cross validation and validation

Usage

  • When using this tool in Python, the result object contains both a feature class and a CrossValidationResult, which has the following properties:

    • Count—Total number of samples used.

    • Mean Error—The averaged difference between the measured and the predicted values.

      Mean error

    • Root Mean Square Error—Indicates how closely your model predicts the measured values. The smaller this error, the better.

      Root mean square error

    • Average Standard Error—The average of the prediction standard errors.

      Average standard error

    • Mean Standardized Error—The average of the standardized errors. This value should be close to 0.

      Mean standardized error

    • Root Mean Square Standardized Error—This should be close to 1 if the prediction standard errors are valid. If the root-mean-squared standardized error is greater than 1, you are underestimating the variability in your predictions. If the root-mean-square-standardized error is less than 1, you are overestimating the variability in your predictions.

      Root mean square standardized error

    • Percent in 90% Interval—The percentage of points that are in a 90 percent cross-validation confidence interval. This value should be close to 90.

    • Percent in 95% Interval—The percentage of points that are in a 95 percent cross-validation confidence interval. This value should be close to 95.

    • Average CRPS—The average Continuous Ranked Probability Score (CRPS) of all points. The CRPS is a diagnostic that measures the deviation from the predictive cumulative distribution function to each observed data value. This value should be as small as possible. This diagnostic has advantages over other cross-validation diagnostics because it compares the data to a full distribution rather than to single-point predictions. The calculation of this statistic involves simulations so it cannot be written in a simple formula.

    Only the Mean and Root Mean Square Error results are available for IDW, Global Polynomial Interpolation, Radial Basis Functions, Diffusion Interpolation With Barriers, and Kernel Interpolation With Barriers.

    Percent in 90% Interval, Percent in 95% Interval, and Average CRPS are only available for Empirical Bayesian Kriging and EBK Regression Prediction models.

  • The fields in the optional output feature class are described in the GA Layer To Points tool.

Parameters

Label Explanation Data type

Input geostatistical layer

The geostatistical layer to be analyzed.

Geostatistical Layer

Output point feature class

(Optional)

Stores the cross-validation statistics at each location in the geostatistical layer.

Feature Class

Derived output

Label Explanation Data type

Count

Total number of samples used.

Long

Mean error

Mean Error—The averaged difference between the measured and the predicted values.

Double

Root mean square

Root Mean Square Error—Indicates how closely your model predicts the measured values.

Double

Average standard

Average Standard Error—The average of the prediction standard errors.

Double

Mean standardized

Mean Standardized Error—The average of the standardized errors.

Double

Root mean square standardized

Root Mean Square Standardized Error—This should be close to 1 if the prediction standard errors are valid.

Double

Percent in 90% Interval

Percent in 90% Interval—The percentage of points that are in a 90 percent cross-validation confidence interval. This value should be close to 90.

Double

Percent in 95% Interval

Percent in 95% Interval—The percentage of points that are in a 95 percent cross-validation confidence interval. This value should be close to 95.

Double

Average CRPS

Average CRPS—The average Continuous Ranked Probability Score (CRPS) of all points. The CRPS is a diagnostic that measures the deviation from the predictive cumulative distribution function to each observed data value. This value should be as small as possible. This diagnostic has advantages over other cross-validation diagnostics because it compares the data to a full distribution rather than to single-point predictions. The calculation of this statistic involves simulations so it cannot be written in a simple formula.

Double

Environments

Current Workspace, Geographic Transformations, Output Coordinate System, Parallel Processing Factor

Licensing information

  • Basic: Requires Geostatistical Analyst
  • Standard: Requires Geostatistical Analyst
  • Advanced: Requires Geostatistical Analyst