Training

Cleaning (or scrubbing) data is one of the most important steps in data analysis, but it is also one of the most neglected. Data cleaning can be tedious and time-consuming, and in some cases, the analyst may not even be aware that it should be done. Proper data cleaning, however, should be the first step in any analysis process, to ensure accurate results.

The new SPSS Data Validation module enables you to streamline the data cleaning process by using rules to perform a variety of data checks. For example, you can specify data validation rules for individual variables, such as range checks, and cross-variable checks (e.g., "pregnant males"), and then store these checks in an SPSS data dictionary to use in future applications.

Let's say, for example, that you want to identify respondents who indicate that they watch more than 10 hours of television per day. These data points are considered to be outliers.

• From the menus, click Data…Validation…Validate Data
• In the Variables tab, highlight Tvhours and move it over the Analysis Variables box
• Highlight ID and move it over the Case Identifier Variables box (see Figure 1)

Figure 1: The Variables screen within the Validate Data dialog box
(Click to enlarge)

• Click the Single-Variables Rules tab (see Figure 2)
• Click Define Rules…

Figure 2: The Single-Variable Rules screen within the Validate Data dialog box (Click to enlarge)

• Change the Minimum values text box to 0 and the Maximum values text box to 10 (See Figure 3)

Figure 3: Select ranges for valid values in the Validate Data dialog box
(Click to enlarge)

• Click Continue
• Click Apply in the Rules Category box
• Click OK

The table in Figure 4 below presents the 12 respondents who indicated that they watch television more than 10 hours per day, and includes the ID needed to examine each record.

Figure 4: The output from the data validation procedure

This is just one example of how the SPSS Data Validation module can help you identify potential errors in your data before you run statistical analyses.