Declustering

If the Need Declustering toggle has been selected in the Grade to be Estimated step, the Declustering Data application is the next step.

The principle of the Declustering application is to assign a weight to each sample where a given variable is defined taking possible clusters of samples into account. To compute the weight wi attached to a target sample i, the program counts the number of samples ni inside a moving window centered on this target. The weight wi is equal to mv/ni where mv is the mean of all the ni.

This application enables the user to choose the optimal Moving Window Size used in the Moving Window Declustering method.

Note: The weight of a sample will be 1 when the number of points inside the Moving Window equals mv the mean of the ni. A low weight will be attached to a sample which is inside a cluster while an isolated point will get the maximum weight.

• Declustering Window Sizes section

  • Enter the Minimum Declustering Window Sizes
  • Enter the Maximum Declustering Window Sizes
  • Enter the Number of moving windows N

Click on Update to update the graph when Declustering Window Sizes parameters have been modified.

Basic statistics (Weighted Mean and Weighted Standard Deviation) are computed for different moving window sizes derived from Minimum Window Size, Maximum Window Size and N parameters. The Raw Mean corresponds to the mean of the test variable using non-weighted samples.

• Test Variable section

  • Select the Test Variable used to compute the Raw Mean, Expected Mean, Weighted Mean and the Weighted Standard Deviation. The available Test Variables are the different grade variables of the corresponding estimation group.

    In a heterotopic case it is advised to choose the test variable that is defined on the maximum number of samples to select a meaningful moving window size. In the below table are displayed the number of samples on which each variable is defined.

  • Expected Mean: It is by default a kriging mean computed on panels inside the domain (selected in the previous step) using a linear model and a moving neighborhood which size is deduced from the panels size.

• Window Size Choice section

  • Size: It corresponds to the optimal size for the Moving Window. Select the suitable declustering size using the combo-box Size or by clicking on a diamond in the graph.

  Note: The closer the Weighted Mean to the Expected mean, the more suitable the Window Size.

• Graphic Option section

  • The display of the Weighted Standard Deviation Curve is optional.
  • Need Declustering (Confirm): Select this option if, after analysis, declustering is required.

Click on to set default values.

Click Next to go to the Data Analysis step.