Checking for a bivariate normal distribution

Available with Geostatistical Analyst license.

This is a means to determine if the data comes from a bivariate normal distribution. The two lines in the Examine Bivariate Distribution dialog box should be close to each other if the data comes from a bivariate normal distribution.

Disjunctive kriging requires the data to come from a bivariate normal distribution. Simple kriging requires that the data come from a multivariate normal distribution, and the bivariate check can help justify the multivariate assumption.

Learn about bivariate normal distributions

1. Click the Add Data button on the ArcMap toolbar and add the layer you want to check for bivariate distribution.
2. Click the Geostatistical Wizard button on the Geostatistical Analyst toolbar.
3. Choose Kriging/Cokriging, specify values for Source Dataset and Data Field, then click Next.
4. Click Simple or Disjunctive as the kriging method and click Next.
5. Click Next on the Normal Score Transformation dialog box.
6. Set the Examine bivariate distribution flag to True and click Next.
7. Enter Quantile values and check that the bivariate and model curves shown in the graph are similar (overlap as much as possible).

   If you are not satisfied with the overlap, click Back and modify the semivariogram or covariance model. You should try to get the bivariate and model curves to match for as many quantiles as possible (evaluate the overlap for quantiles of 0.1 to 0.9 with increments of 0.1, for example).

8. When satisfied with the overlap, click Next.
9. On the Searching Neighborhood dialog box, click Next.
10. Examine the results on the Cross Validation dialog box and click Finish.
11. Click OK on the Method Report dialog box to create the surface.