Removing global and local trends from the data: Detrending
Use the Explore Data trend analysis tool to determine if there is a global trend in the data.
Basically, you are decomposing your data into a deterministic trend component and an autocorrelated random component. Once the trend is removed, you will carry out kriging on the residuals. Before the final predictions are actually calculated, the trend is added back to the output surface.
手順:
- Select either Ordinary, Universal, or Disjunctive kriging and the desired output surface type on the Geostatistical Wizard dialog box.
- Click the Order of Trend Removal drop-down arrow, click the desired order for the trend, then click Next.
- Increase the value on the Exploratory Trend Surface Analysis slider to define the window size for fitting the polynomial from Global to Local.
- Click Next.
- Specify the desired parameters on the Semivariogram/Covariance Modeling dialog box and click Next.
- Specify the desired parameters on the Searching Neighborhood dialog box and click Next.
- Examine the results on the Cross Validation dialog box and click Finish.
- Click OK on the Method Report dialog box.
Alternatively, set the Advanced mode parameter to True to change the neighborhood parameters.
ヒント:
Keep your trends as low-order Global Polynomials unless you have good reason to do otherwise.
7/10/2012