Modeling semivariogram and covariance functions: Exploring directional autocorrelation

Directional autocorrelation may exist in your data and can be accounted for by incorporating anisotropy in the semivariogram model.

As you perform the steps below, watch the scatter of the points in the semivariogram around the model. If the points are close to the line in one direction and are spread out in another, there may be a directional autocorrelation in your data. The exploration for directional autocorrelation using the search direction, in conjunction with the directional indicator in the semivariogram map, is only graphic (around the isotropic fitted model). To fit a model with directional autocorrelation, check the Anisotropy box.

Etapes :
  1. Set Show search direction in the View Settings panel to True.
  2. Click the up or down arrow of the Angle control to explore for the desired directional angle.
    • Alternatively, in the semivariogram map, click and hold the middle blue line and drag in the desired direction.
  3. Click the up or down arrow of the Tolerance control to adjust the angle of tolerance.
    • Alternatively, in the semivariogram map, click and hold either of the red directional indicator lines and drag to the desired angle.
  4. Click the up or down arrow of the Bandwidth control.
    • Alternatively, in the semivariogram map, click and hold either of the red lines located on the bounding square and drag to the desired width.
RemarqueRemarque :

Be aware that even if the Show search direction parameter indicates a specific direction, the semivariogram is constructed based on the pairs of points that are within the specified direction and those that are located in the opposite one. With this, the tool is "symmetrical." When the indicated angle is, for example, 45°, it means that it checks both the 45° and 45+180=225°.

4/26/2014