## Extracting Observations

You can compare analyses made with different observations by
*extracting* a subset, that is, by creating a new data set
that contains a subset of observations from the original data set.
Then you can request separate analyses for each data set.
Consider the **MINING** data. This data set contains results
of an experiment to examine drilling times (**DRILTIME**)
for different drilling methods (**METHOD**).
As it turned out, the experimenters encountered difficulties
due to changing rock types after a depth of about 200 feet.
It might be worthwhile to compare the distribution of
**DRILTIME** for depths greater than 200 feet to the
distribution of **DRILTIME** for the entire data set.
To compare the two distributions, you need to select the
observations where **DEPTH** is greater than 200 feet and
extract them into a new data window.
| Open the **MINING** data set. |

**Figure 21.2:** **MINING** Data

| Choose **Edit:Observations:Find**. |

**Figure 21.3:** Finding Observations

This displays the **Find Observations** dialog.

**Figure 21.4:** Find Observations Dialog

| Select **>** in the **Test** list and **200** in the **Value** list. |

**Figure 21.5:** Finding **DEPTH** > 200

| Click the **OK** button. |

This selects all observations where **DEPTH** is greater than
200 feet. To see the selected observations, either choose
**Find Next** from the data pop-up menu or scroll down using
the vertical scroll bar on the right (as indicated by the
arrow in the figure).

**Figure 21.6:** Observations where **DEPTH** > 200

| Choose **Extract** from the data pop-up menu. |

A new data set containing observations where **DEPTH** is greater
than 200 feet appears, as shown in Figure 21.7.
The new data window is named automatically
by adding a subscript to the original name.
You may have to scroll to the top of the data window
to duplicate the next figure.

**Figure 21.7:** **MINING1** Data

Now create distribution analyses for both data sets.

| Select **DRILTIME** in the **MINING** data window. |

| Choose **Analyze:Distribution (Y)**. |

A distribution analysis using
all the observations appears on your display.

| Select **DRILTIME** in the **MINING1** data window. |

| Choose **Analyze:Distribution (Y)**. |

A distribution analysis using the subset
of observations appears on your display.

| Move the two analysis windows side-by-side to compare the distributions. |

**Figure 21.8:** Comparing Two Distribution Analyses

The mean drilling time at depths greater than 200 feet
was 9.9601, while the mean overall was only 8.8589.
The drills may have found harder rock at greater depths.
You may want to create an additional analysis to compare depths
greater than 200 feet with depths less than or equal to 200 feet.

| Choose **File:End** to delete **MINING1** and the two analysis windows. |

Note | Sometimes you will want to compare analyses that use different subsets of
observations based on the values of some variable.
If this is the case, you can assign the variable the
**Group** role, as described in Chapter 22, "Analyzing by Groups." |

Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.