cepp_logo5.gif (13296 bytes)


newspaper.gif (867 bytes)

<Back to Articles Page

The following article was written by Coleman Patterson and appeared in the Business section of the Abilene Reporter-News.

Good information comes from goo surveys, May 18, 2008, 2D.

Since problem solving is one of the major elements of a manager’s job, having access to timely and accurate business information is critically important.  Information can come from formal and informal organizational communication and from sales, financial, economic, and industry data.  Managers might also find it beneficial to collect and analyze information about the attitudes of workers and consumers using surveys.

When properly constructed and analyzed, surveys can provide managers with valuable information.  There is a science behind good survey construction.  Error and threats to validity can creep into the survey process at many different points.  Improperly worded or confusing questions, selection of a non-representative sample, data entry errors, and measurement errors can all affect the validity, or accuracy, of the survey results.  One common problem encountered by novice survey writers is the over-use of questions using nominal-level data.

Nominal data, or named data, are numbers that represent different categories or groups of data.  Gender, which is a commonly asked survey question, is an example of nominal-level data.  The numbers entered into the data set for gender are simply codes for male and female.  It is common to code variables like gender using zeros and ones or ones and twos.  However, since the numbers are merely codes for the categories, any numbers could be used—such as 1,000 for males and 3.14 for females.   Asking respondents to “check all that apply” and “identify your favorite choice” typically result in nominal data.

Nominal data can also be created from higher-level data by breaking a variable into groups.  Annual income, for example, is frequently asked in pay ranges on surveys.  It is also common to ask respondents to indicate the number of visits or number of purchases made over a specified amount of time using artificially developed groups.  Less than five, five to ten, or more than ten are examples of ways to break real numbers into groups.  However, those same variables could also be divided in a multitude of other ways.  Being nominal data, codes have to be assigned to the categories and entered into the data set.

Once coded and recorded in the data set, the options for describing and analyzing nominal data are fairly limited.  For example, it does not make sense to calculate averages of the coded numbers for nominal data variables.  Averages and other higher-level descriptive statistics only make sense for variables that are real numbers—such as the number of products sold, employee pay, amount of inventory, advertising expenditures, etc.  The appropriate statistical analyses for nominal data are ones that count and report the numbers of observations in each group.  Researchers can statistically analyze differences in the counts or proportions of observations in each group.  

When combined with variables that are real numbers, nominal data can be used in more sophisticated statistical analyses—for example, is there is difference in sales between male and female salespeople?  Good and useful information comes from good surveys.  Decision makers should invest time and attention learning proper survey construction techniques.

<Back to Articles Page

reporternews.gif (5314 bytes)

© 2006, 2007, 2008  Coleman Patterson, All Rights Reserved