Deciding how many classes to use for grouping the data is a compromise between the extremes of too much detail (each observation in its own category) and not enough detail (only one category). Most frequency tables are constructed according to the following guidelines:
- For most data, 6 to 15 classes are enough
- Class intervals (lengths) should be equal. Intervals such as 0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50, etc are desirable.
- The starting point for each class should be divisible by the interval, For example, in the class 15 - 20, the starting point, 15, is divisible by the interval, 5.
- Each observation must fit into only one class.
- When a large number of points falls around a certain value, make this value the approximate center of the frequency distribution.
For the data in our example, the minimum is 65 and the maximum is 114, a range of about 50. We can therefore choose intervals of size 5, and have ten of them. Our classes are 65 - 70, 70 - 75, etc.