Previous Page  Next Page 
Comparing Analyses

Excluding Observations

Another way to compare analyses using different observations is to exclude observations, that is, to remove them from calculations in the analysis. The observations still appear in graphs. To illustrate this technique, consider a simple linear regression model with DRILTIME as the response variable and DEPTH as the explanatory variable.

Select DRILTIME, then DEPTH, then choose Analyze:Fit (Y X).

This displays a fit window.

com09.gif (16814 bytes)

Figure 21.9: Fit Window

Choose Edit:Windows:Copy Window in the fit window.

This creates a copy of the fit window.

Figure 21.10: Edit:Windows Menu

Move the two fit windows side by side.

Choose Edit:Windows:Freeze in the fit window on the left.

Figure 21.11: Edit:Windows Menu

This freezes the window, as indicated by the frost in the corners of the window. Freezing a window converts the window to a static image that ignores any changes to the data. Normally, all SAS/INSIGHT windows are linked to their data, and any changes to the data are automatically reflected in all analyses. By freezing a window, you can compare windows using different observations without creating additional data sets.

com12.gif (30268 bytes)

Figure 21.12: Two Windows, One Frozen

Now exclude a few observations from the window on the right.

Select the three observations with the largest values of DRILTIME in the scatter plot.

Choose Edit:Observations:Exclude in Calculations.

Figure 21.13: Edit: Observations Menu

This recalculates the fit analysis without the selected observations. Normally, both windows would be recalculated, but since the window on the left is frozen, it does not change. Now you can compare the two fit windows.

com14.gif (31062 bytes)

Figure 21.14: Comparing Two Fit Windows

To thaw a frozen window, follow these steps.

Choose Edit:Windows:Freeze again.

This recalculates the frozen window and restores its dynamic behavior.

Close all analysis windows before proceeding to the next section.

Previous Page  Next Page  Top of Page

Copyright © 2007 by SAS Institute Inc., Cary, NC, USA. All rights reserved.