Discovering coding issues using frequencies

The frequency distribution for INCOME in the following screenshot demonstrates another reason why it is important to examine the pattern for individual data fields before diving into analytics more deeply. Navigate to Analyze | Descriptive Statistics | Frequencies, and select Respondents Income to build this table:

People are often reluctant to divulge their income so surveys typically ask them to select an income category like the groupings in this table. Notice, however, that the groups (numbered 1 through 26) represent unequal bands of income. The groups coded 3 thru 7 represent a range of only $1,000, while group 15 has a range of $5,000, and group 20 covers $15,000.

In the previous chapter, one of the properties of variables that was discussed involved the level of measurement. Income is an example of a scale variable and for these fields, it is assumed that the intervals between values are equal, which is not the case in this instance. The information in this table can be evaluated even though the categories are not uniform in terms of the dollar range they represent, but before using this field for statistical modeling, it is important to adjust the coding to reflect the actual differences in income.

In a later chapter, techniques for modifying fields will be covered but it is by examining the data along with the coding scheme used that you discover where the need for change exists.