Calculating Statistics Using Intervals

When analyzed, write-in number questions display only calculated values, not a distribution as with multiple choice questions. It is difficult to get a distribution of the data although it can be done by recoding values using Tools.

If it is important to get both an average and a distribution, use a weighted score instead of write-in number. Use intervals that have a midpoint and use the midpoint as the code. The calculated values will then be reasonably good approximations of wha t would be obtained by a write-in number.

To test this idea, start a new project with these two questions. We will assume that the survey is being given to adults that are all over 5 feet (60 inches) tall. Basketball players?

What is your height in inches? (Write-in number type question.)

Select the interval that includes your height in inches to the nearest inch. (Weighted response type question)

Question Interval Code

60 64 62

65 69 67

70 74 72

75 79 77

80 84 82

85 89 87

90 94 92

Try it out by recording at least 20 cases with random values. To get a random sample you could ask members of the class to write down two or three numbers between 60 and 94 on a slip of paper and give it to you. Remember that the interval has to agree with the write-in number. If you put in the height at 68 inches, you must pick the interval 65 69.

Do an analysis and see if the average as calculated by the write-in number (to the nearest inch) agrees with the average as calculated with the intervals.

This method assumes that the average height of people in each interval is the same as the code number. This is usually a pretty good assumption, at least for a random sample of a population. This method is appropriate for such demographic variables as height, weight, age, income, etc.